From routing-discussion-admin@ietf.org  Tue Jul  2 04:28:41 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA28473
	for <routing-discussion-archive@odin.ietf.org>; Tue, 2 Jul 2002 04:28:41 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id EAA21543;
	Tue, 2 Jul 2002 04:17:12 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id EAA21514
	for <routing-discussion@optimus.ietf.org>; Tue, 2 Jul 2002 04:17:10 -0400 (EDT)
Received: from gorilla.mchh.siemens.de (gorilla.mchh.siemens.de [194.138.158.18])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA28310
	for <routing-discussion@ietf.org>; Tue, 2 Jul 2002 04:16:22 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by gorilla.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id KAA14490
	for <routing-discussion@ietf.org>; Tue, 2 Jul 2002 10:17:07 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id KAA14595
	for <routing-discussion@ietf.org>; Tue, 2 Jul 2002 10:17:07 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2653.19)
	id <MYNDYLZX>; Tue, 2 Jul 2002 10:17:07 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B8D8@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: routing-discussion@ietf.org
Subject: Re: Loop Detection
Date: Tue, 2 Jul 2002 10:16:59 +0200 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Wouldn't it be a fine goal to standardize a tool/mechanism/protocol
which double-checks, within milliseconds, that all routing tables of some network domain are in sync ?
Or at least, that the topological view, aquired by all nodes of some network domain complies with the physical reality ?
Wouldn't this be a powerful OAM-application ? IMHO:it would be better than the religious fights on OAM-MPLS.

Heinrich Hummel
Siemens AG



 On Tue, 4 Jun 2002, Jing Shen wrote:

> Given a network with several routers, is there any tool/application to
> detect loops in the routing tables?

As far as I know, there isn't.
Also if the loop is due to a misconfiguration which does not show up
in BGP updates, then it's harder to detect.

-Sue


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Wed Jul  3 18:20:44 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA09982
	for <routing-discussion-archive@odin.ietf.org>; Wed, 3 Jul 2002 18:20:44 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id SAA18416;
	Wed, 3 Jul 2002 18:13:40 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id SAA18388
	for <routing-discussion@ns.ietf.org>; Wed, 3 Jul 2002 18:13:38 -0400 (EDT)
Received: from smtp.cwnt.com (smtp.cwnt.com [192.116.246.129])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA09402
	for <routing-discussion@ietf.org>; Wed, 3 Jul 2002 18:12:09 -0400 (EDT)
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Subject: RE: Loop Detection
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
Date: Thu, 4 Jul 2002 01:11:53 +0300
Message-ID: <C7DF4400240AFB4095D98C7C6EC2A34A73012B@bart.cwnt.com>
Thread-Topic: Loop Detection
Thread-Index: AcIhoRADvL9Q4awOQBquVHPqESNSyABO+NYg
From: "Amir Hermelin" <amir@cwnt.com>
To: "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id SAA18389
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

my 2 cents:
* Yes, such a tool would be very nice. Although, a small portion of our salaries is due to the fact that routing loops are so scary ;-)
* The approaches would be different for types of protocols. In IGPs, assuming correct implementations, the protocol takes care of loop possibilities as long as the lspdb is in sync. When its not due to transient states, a tool wouldn't help since you'd have to run it constantly for every router's view.
* I agree that detecting BGP mis-configuration is hard. As for MPLS, there are methods that are more susceptible to loops than others (e.g. one implementation of TE-shortcuts can lead to loops even in a seemingly correct configuration, due to metric values).
* If this stuff moves forward, it might even mandate a whole new workgroup. 

--
Amir Hermelin                  <mailto:amir@cwnt.com>
Charlotte's Web Networks Inc.   <http://www.cwnt.com>


> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Tuesday, July 02, 2002 11:17 AM
> To: routing-discussion@ietf.org
> Subject: Re: Loop Detection
> 
> 
> Wouldn't it be a fine goal to standardize a tool/mechanism/protocol
> which double-checks, within milliseconds, that all routing 
> tables of some network domain are in sync ?
> Or at least, that the topological view, aquired by all nodes 
> of some network domain complies with the physical reality ?
> Wouldn't this be a powerful OAM-application ? IMHO:it would 
> be better than the religious fights on OAM-MPLS.
> 
> Heinrich Hummel
> Siemens AG
> 
> 
> 
>  On Tue, 4 Jun 2002, Jing Shen wrote:
> 
> > Given a network with several routers, is there any 
> tool/application to
> > detect loops in the routing tables?
> 
> As far as I know, there isn't.
> Also if the loop is due to a misconfiguration which does not show up
> in BGP updates, then it's harder to detect.
> 
> -Sue
> 
> 
> _______________________________________________
> routing-discussion mailing list
> routing-discussion@ietf.org
> https://www1.ietf.org/mailman/listinfo/routing-discussion
> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Wed Jul  3 19:03:33 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA11811
	for <routing-discussion-archive@odin.ietf.org>; Wed, 3 Jul 2002 19:03:33 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id SAA00600;
	Wed, 3 Jul 2002 18:59:48 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id SAA00572
	for <routing-discussion@ns.ietf.org>; Wed, 3 Jul 2002 18:59:45 -0400 (EDT)
Received: from mail6.mn.rr.com (fe6.rdc-kc.rr.com [24.94.163.53])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA11577
	for <routing-discussion@ietf.org>; Wed, 3 Jul 2002 18:58:55 -0400 (EDT)
Received: from jrenwick ([24.26.184.20]) by mail6.mn.rr.com  with Microsoft SMTPSVC(5.5.1877.537.53);
	 Wed, 3 Jul 2002 17:56:25 -0500
From: "John Renwick" <jkr@visi.com>
To: "Amir Hermelin" <amir@cwnt.com>,
        "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
Subject: RE: Loop Detection
Date: Wed, 3 Jul 2002 17:59:07 -0500
Message-ID: <PAEBIODDHHIMCKLJKMADCEBPDAAA.jkr@visi.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2911.0)
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Importance: Normal
In-Reply-To: <C7DF4400240AFB4095D98C7C6EC2A34A73012B@bart.cwnt.com>
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Isn't TTL a very direct and immediate form of loop detection?  
If you had a network tool that would illuminate the routers that 
are counting a lot of packets dropped because of TTL expiration, 
it could be showing you points on a routing loop.

--
John Renwick                          Agere Systems
Software Technical Manager            1230 Northland Drive
Phone: (651) 675-2986                 Mendota Heights, MN 55120
FAX:   (707) 897-7928                 Cell:  (612) 669-9397
mailto:jrenwick@agere.com             http://www.agere.com

> -----Original Message-----
> From: routing-discussion-admin@ietf.org
> [mailto:routing-discussion-admin@ietf.org]On Behalf Of Amir Hermelin
> Sent: Wednesday, July 03, 2002 5:12 PM
> To: Hummel Heinrich; routing-discussion@ietf.org
> Subject: RE: Loop Detection
> 
> 
> my 2 cents:
> * Yes, such a tool would be very nice. Although, a small portion 
> of our salaries is due to the fact that routing loops are so scary ;-)
> * The approaches would be different for types of protocols. In 
> IGPs, assuming correct implementations, the protocol takes care 
> of loop possibilities as long as the lspdb is in sync. When its 
> not due to transient states, a tool wouldn't help since you'd 
> have to run it constantly for every router's view.
> * I agree that detecting BGP mis-configuration is hard. As for 
> MPLS, there are methods that are more susceptible to loops than 
> others (e.g. one implementation of TE-shortcuts can lead to loops 
> even in a seemingly correct configuration, due to metric values).
> * If this stuff moves forward, it might even mandate a whole new 
> workgroup. 
> 
> --
> Amir Hermelin                  <mailto:amir@cwnt.com>
> Charlotte's Web Networks Inc.   <http://www.cwnt.com>
> 
> 
> > -----Original Message-----
> > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> > Sent: Tuesday, July 02, 2002 11:17 AM
> > To: routing-discussion@ietf.org
> > Subject: Re: Loop Detection
> > 
> > 
> > Wouldn't it be a fine goal to standardize a tool/mechanism/protocol
> > which double-checks, within milliseconds, that all routing 
> > tables of some network domain are in sync ?
> > Or at least, that the topological view, aquired by all nodes 
> > of some network domain complies with the physical reality ?
> > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > be better than the religious fights on OAM-MPLS.
> > 
> > Heinrich Hummel
> > Siemens AG
> > 
> > 
> > 
> >  On Tue, 4 Jun 2002, Jing Shen wrote:
> > 
> > > Given a network with several routers, is there any 
> > tool/application to
> > > detect loops in the routing tables?
> > 
> > As far as I know, there isn't.
> > Also if the loop is due to a misconfiguration which does not show up
> > in BGP updates, then it's harder to detect.
> > 
> > -Sue
> > 
> > 
> > _______________________________________________
> > routing-discussion mailing list
> > routing-discussion@ietf.org
> > https://www1.ietf.org/mailman/listinfo/routing-discussion
> > 
> 
> _______________________________________________
> routing-discussion mailing list
> routing-discussion@ietf.org
> https://www1.ietf.org/mailman/listinfo/routing-discussion
> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Thu Jul  4 09:18:06 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA16829
	for <routing-discussion-archive@odin.ietf.org>; Thu, 4 Jul 2002 09:18:06 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id JAA18175;
	Thu, 4 Jul 2002 09:11:08 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id JAA18145
	for <routing-discussion@optimus.ietf.org>; Thu, 4 Jul 2002 09:11:06 -0400 (EDT)
Received: from zcars04e.ca.nortel.com (zcars04e.nortelnetworks.com [47.129.242.56])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA16711
	for <routing-discussion@ietf.org>; Thu, 4 Jul 2002 09:10:15 -0400 (EDT)
Received: from zcard015.ca.nortel.com (zcard015.ca.nortel.com [47.129.30.7])
	by zcars04e.ca.nortel.com (Switch-2.2.0/Switch-2.2.0) with ESMTP id g64DASk08831;
	Thu, 4 Jul 2002 09:10:28 -0400 (EDT)
Received: by zcard015.ca.nortel.com with Internet Mail Service (5.5.2653.19)
	id <NYVB0YKQ>; Thu, 4 Jul 2002 09:10:28 -0400
Message-ID: <9FBD322B7824D511B36900508BF93C9C0476C664@zcard031.ca.nortel.com>
From: "David Allan" <dallan@nortelnetworks.com>
To: John Renwick <jkr@visi.com>, Amir Hermelin <amir@cwnt.com>,
        Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: RE: Loop Detection
Date: Thu, 4 Jul 2002 09:10:27 -0400 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: multipart/alternative;
	boundary="----_=_NextPart_001_01C2235C.2A1E60B4"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_001_01C2235C.2A1E60B4
Content-Type: text/plain;
	charset="iso-8859-1"

Issue is it only tells you ***somthing*** is broken.

Similarly simply ensuring that the routing databases are synchronized does
not perform a sanity check on the local derivative by-products (FIBs, FTN,
NHLFE etc.) or the hw that implements them.

cheers
Dave

> -----Original Message-----
> From: John Renwick [mailto:jkr@visi.com]
> Sent: Wednesday, July 03, 2002 6:59 PM
> To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org
> Subject: RE: Loop Detection
> 
> 
> Isn't TTL a very direct and immediate form of loop detection?  
> If you had a network tool that would illuminate the routers that 
> are counting a lot of packets dropped because of TTL expiration, 
> it could be showing you points on a routing loop.
> 
> --
> John Renwick                          Agere Systems
> Software Technical Manager            1230 Northland Drive
> Phone: (651) 675-2986                 Mendota Heights, MN 55120
> FAX:   (707) 897-7928                 Cell:  (612) 669-9397
> mailto:jrenwick@agere.com             http://www.agere.com
> 
> > -----Original Message-----
> > From: routing-discussion-admin@ietf.org
> > [mailto:routing-discussion-admin@ietf.org]On Behalf Of Amir Hermelin
> > Sent: Wednesday, July 03, 2002 5:12 PM
> > To: Hummel Heinrich; routing-discussion@ietf.org
> > Subject: RE: Loop Detection
> > 
> > 
> > my 2 cents:
> > * Yes, such a tool would be very nice. Although, a small portion 
> > of our salaries is due to the fact that routing loops are 
> so scary ;-)
> > * The approaches would be different for types of protocols. In 
> > IGPs, assuming correct implementations, the protocol takes care 
> > of loop possibilities as long as the lspdb is in sync. When its 
> > not due to transient states, a tool wouldn't help since you'd 
> > have to run it constantly for every router's view.
> > * I agree that detecting BGP mis-configuration is hard. As for 
> > MPLS, there are methods that are more susceptible to loops than 
> > others (e.g. one implementation of TE-shortcuts can lead to loops 
> > even in a seemingly correct configuration, due to metric values).
> > * If this stuff moves forward, it might even mandate a whole new 
> > workgroup. 
> > 
> > --
> > Amir Hermelin                  <mailto:amir@cwnt.com>
> > Charlotte's Web Networks Inc.   <http://www.cwnt.com>
> > 
> > 
> > > -----Original Message-----
> > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> > > Sent: Tuesday, July 02, 2002 11:17 AM
> > > To: routing-discussion@ietf.org
> > > Subject: Re: Loop Detection
> > > 
> > > 
> > > Wouldn't it be a fine goal to standardize a 
> tool/mechanism/protocol
> > > which double-checks, within milliseconds, that all routing 
> > > tables of some network domain are in sync ?
> > > Or at least, that the topological view, aquired by all nodes 
> > > of some network domain complies with the physical reality ?
> > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > be better than the religious fights on OAM-MPLS.
> > > 
> > > Heinrich Hummel
> > > Siemens AG
> > > 
> > > 
> > > 
> > >  On Tue, 4 Jun 2002, Jing Shen wrote:
> > > 
> > > > Given a network with several routers, is there any 
> > > tool/application to
> > > > detect loops in the routing tables?
> > > 
> > > As far as I know, there isn't.
> > > Also if the loop is due to a misconfiguration which does 
> not show up
> > > in BGP updates, then it's harder to detect.
> > > 
> > > -Sue
> > > 
> > > 
> > > _______________________________________________
> > > routing-discussion mailing list
> > > routing-discussion@ietf.org
> > > https://www1.ietf.org/mailman/listinfo/routing-discussion
> > > 
> > 
> > _______________________________________________
> > routing-discussion mailing list
> > routing-discussion@ietf.org
> > https://www1.ietf.org/mailman/listinfo/routing-discussion
> > 
> 
> _______________________________________________
> routing-discussion mailing list
> routing-discussion@ietf.org
> https://www1.ietf.org/mailman/listinfo/routing-discussion
> 

------_=_NextPart_001_01C2235C.2A1E60B4
Content-Type: text/html;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Diso-8859-1">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
5.5.2655.35">
<TITLE>RE: Loop Detection</TITLE>
</HEAD>
<BODY>

<P><FONT SIZE=3D2>Issue is it only tells you ***somthing*** is =
broken.</FONT>
</P>

<P><FONT SIZE=3D2>Similarly simply ensuring that the routing databases =
are synchronized does not perform a sanity check on the local =
derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that =
implements them.</FONT></P>

<P><FONT SIZE=3D2>cheers</FONT>
<BR><FONT SIZE=3D2>Dave</FONT>
</P>

<P><FONT SIZE=3D2>&gt; -----Original Message-----</FONT>
<BR><FONT SIZE=3D2>&gt; From: John Renwick [<A =
HREF=3D"mailto:jkr@visi.com">mailto:jkr@visi.com</A>]</FONT>
<BR><FONT SIZE=3D2>&gt; Sent: Wednesday, July 03, 2002 6:59 PM</FONT>
<BR><FONT SIZE=3D2>&gt; To: Amir Hermelin; Hummel Heinrich; =
routing-discussion@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; Subject: RE: Loop Detection</FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; Isn't TTL a very direct and immediate form of =
loop detection?&nbsp; </FONT>
<BR><FONT SIZE=3D2>&gt; If you had a network tool that would illuminate =
the routers that </FONT>
<BR><FONT SIZE=3D2>&gt; are counting a lot of packets dropped because =
of TTL expiration, </FONT>
<BR><FONT SIZE=3D2>&gt; it could be showing you points on a routing =
loop.</FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; --</FONT>
<BR><FONT SIZE=3D2>&gt; John =
Renwick&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
;&nbsp;&nbsp; Agere Systems</FONT>
<BR><FONT SIZE=3D2>&gt; Software Technical =
Manager&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
; 1230 Northland Drive</FONT>
<BR><FONT SIZE=3D2>&gt; Phone: (651) =
675-2986&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Mendota Heights, MN 55120</FONT>
<BR><FONT SIZE=3D2>&gt; FAX:&nbsp;&nbsp; (707) =
897-7928&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Cell:&nbsp; (612) 669-9397</FONT>
<BR><FONT SIZE=3D2>&gt; <A =
HREF=3D"mailto:jrenwick@agere.com">mailto:jrenwick@agere.com</A>&nbsp;&n=
bsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; <A =
HREF=3D"http://www.agere.com" =
TARGET=3D"_blank">http://www.agere.com</A></FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; -----Original Message-----</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; From: =
routing-discussion-admin@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; [<A =
HREF=3D"mailto:routing-discussion-admin@ietf.org">mailto:routing-discuss=
ion-admin@ietf.org</A>]On Behalf Of Amir Hermelin</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Sent: Wednesday, July 03, 2002 5:12 =
PM</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; To: Hummel Heinrich; =
routing-discussion@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Subject: RE: Loop Detection</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; my 2 cents:</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; * Yes, such a tool would be very nice. =
Although, a small portion </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; of our salaries is due to the fact that =
routing loops are </FONT>
<BR><FONT SIZE=3D2>&gt; so scary ;-)</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; * The approaches would be different for =
types of protocols. In </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; IGPs, assuming correct implementations, =
the protocol takes care </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; of loop possibilities as long as the lspdb =
is in sync. When its </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; not due to transient states, a tool =
wouldn't help since you'd </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; have to run it constantly for every =
router's view.</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; * I agree that detecting BGP =
mis-configuration is hard. As for </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; MPLS, there are methods that are more =
susceptible to loops than </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; others (e.g. one implementation of =
TE-shortcuts can lead to loops </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; even in a seemingly correct configuration, =
due to metric values).</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; * If this stuff moves forward, it might =
even mandate a whole new </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; workgroup. </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; --</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Amir =
Hermelin&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; &lt;<A =
HREF=3D"mailto:amir@cwnt.com">mailto:amir@cwnt.com</A>&gt;</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Charlotte's Web Networks Inc.&nbsp;&nbsp; =
&lt;<A HREF=3D"http://www.cwnt.com" =
TARGET=3D"_blank">http://www.cwnt.com</A>&gt;</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; -----Original Message-----</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; From: Hummel Heinrich [<A =
HREF=3D"mailto:Heinrich.Hummel@icn.siemens.de">mailto:Heinrich.Hummel@ic=
n.siemens.de</A>]</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Sent: Tuesday, July 02, 2002 11:17 =
AM</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; To: =
routing-discussion@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Subject: Re: Loop Detection</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Wouldn't it be a fine goal to =
standardize a </FONT>
<BR><FONT SIZE=3D2>&gt; tool/mechanism/protocol</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; which double-checks, within =
milliseconds, that all routing </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; tables of some network domain are in =
sync ?</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Or at least, that the topological =
view, aquired by all nodes </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; of some network domain complies with =
the physical reality ?</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Wouldn't this be a powerful =
OAM-application ? IMHO:it would </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; be better than the religious fights =
on OAM-MPLS.</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Heinrich Hummel</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Siemens AG</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt;&nbsp; On Tue, 4 Jun 2002, Jing Shen =
wrote:</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Given a network with several =
routers, is there any </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; tool/application to</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; detect loops in the routing =
tables?</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; As far as I know, there isn't.</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Also if the loop is due to a =
misconfiguration which does </FONT>
<BR><FONT SIZE=3D2>&gt; not show up</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; in BGP updates, then it's harder to =
detect.</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; -Sue</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; =
_______________________________________________</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; routing-discussion mailing =
list</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; routing-discussion@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; <A =
HREF=3D"https://www1.ietf.org/mailman/listinfo/routing-discussion" =
TARGET=3D"_blank">https://www1.ietf.org/mailman/listinfo/routing-discuss=
ion</A></FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; =
_______________________________________________</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; routing-discussion mailing list</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; routing-discussion@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; <A =
HREF=3D"https://www1.ietf.org/mailman/listinfo/routing-discussion" =
TARGET=3D"_blank">https://www1.ietf.org/mailman/listinfo/routing-discuss=
ion</A></FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; =
_______________________________________________</FONT>
<BR><FONT SIZE=3D2>&gt; routing-discussion mailing list</FONT>
<BR><FONT SIZE=3D2>&gt; routing-discussion@ietf.org</FONT>
<BR><FONT SIZE=3D2>&gt; <A =
HREF=3D"https://www1.ietf.org/mailman/listinfo/routing-discussion" =
TARGET=3D"_blank">https://www1.ietf.org/mailman/listinfo/routing-discuss=
ion</A></FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
</P>

</BODY>
</HTML>
------_=_NextPart_001_01C2235C.2A1E60B4--

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Mon Jul  8 11:22:38 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA19358
	for <routing-discussion-archive@odin.ietf.org>; Mon, 8 Jul 2002 11:22:38 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id LAA23744;
	Mon, 8 Jul 2002 11:15:10 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id LAA23716
	for <routing-discussion@optimus.ietf.org>; Mon, 8 Jul 2002 11:15:08 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA18879
	for <routing-discussion@ietf.org>; Mon, 8 Jul 2002 11:14:15 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id RAA02362;
	Mon, 8 Jul 2002 17:14:28 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id RAA23887;
	Mon, 8 Jul 2002 17:14:42 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2653.19)
	id <MYN1ACR7>; Mon, 8 Jul 2002 17:14:42 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B8FC@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: routing-discussion@ietf.org, "'jkr@visi.com'" <jkr@visi.com>,
        "'amir@cwnt.com'" <amir@cwnt.com>,
        "'dallan@nortelnetworks.com'"
	 <dallan@nortelnetworks.com>
Subject: Re: Loop Detection
Date: Mon, 8 Jul 2002 17:14:40 +0200 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

It is not about TTL (which is to control a particular IP-packet on its way to the destination).
Its about a  sanity check w.r.t. the distributed view of the intra-domain network topology. Whether or not IP flows are
on the go is something else (they may do TTL). 

This sanity check may be simple and fast. And if you want, there may be different types of sanity checks:
some which controls just the topology without reachability info,and  without link parameters. It will be the fastest
(microseconds). Next there may be slower but more comprising sanity checks. At last there may even be checks on
proper derivatives (Routing Tables, NHLFEs,...)
Heinrich




Title: RE: Loop Detection
Issue is it only tells you ***somthing*** is broken. 

Similarly simply ensuring that the routing databases are synchronized does not perform a sanity check on the local derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that implements them.

cheers 
Dave 

> -----Original Message----- 
> From: John Renwick [mailto:jkr@visi.com] 
> Sent: Wednesday, July 03, 2002 6:59 PM 
> To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org 
> Subject: RE: Loop Detection 
> 
> 
> Isn't TTL a very direct and immediate form of loop detection?  
> If you had a network tool that would illuminate the routers that 
> are counting a lot of packets dropped because of TTL expiration, 
> it could be showing you points on a routing loop. 
> 
> -- 
> John Renwick                          Agere Systems 
> Software Technical Manager            1230 Northland Drive 
> Phone: (651) 675-2986                 Mendota Heights, MN 55120 
> FAX:   (707) 897-7928                 Cell:  (612) 669-9397 
> mailto:jrenwick@agere.com             http://www.agere.com 
> 
> > -----Original Message----- 
> > From: routing-discussion-admin@ietf.org 
> > [mailto:routing-discussion-admin@ietf.org]On Behalf Of Amir Hermelin 
> > Sent: Wednesday, July 03, 2002 5:12 PM 
> > To: Hummel Heinrich; routing-discussion@ietf.org 
> > Subject: RE: Loop Detection 
> > 
> > 
> > my 2 cents: 
> > * Yes, such a tool would be very nice. Although, a small portion 
> > of our salaries is due to the fact that routing loops are 
> so scary ;-) 
> > * The approaches would be different for types of protocols. In 
> > IGPs, assuming correct implementations, the protocol takes care 
> > of loop possibilities as long as the lspdb is in sync. When its 
> > not due to transient states, a tool wouldn't help since you'd 
> > have to run it constantly for every router's view. 
> > * I agree that detecting BGP mis-configuration is hard. As for 
> > MPLS, there are methods that are more susceptible to loops than 
> > others (e.g. one implementation of TE-shortcuts can lead to loops 
> > even in a seemingly correct configuration, due to metric values). 
> > * If this stuff moves forward, it might even mandate a whole new 
> > workgroup. 
> > 
> > -- 
> > Amir Hermelin                  <mailto:amir@cwnt.com> 
> > Charlotte's Web Networks Inc.   <http://www.cwnt.com> 
> > 
> > 
> > > -----Original Message----- 
> > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de] 
> > > Sent: Tuesday, July 02, 2002 11:17 AM 
> > > To: routing-discussion@ietf.org 
> > > Subject: Re: Loop Detection 
> > > 
> > > 
> > > Wouldn't it be a fine goal to standardize a 
> tool/mechanism/protocol 
> > > which double-checks, within milliseconds, that all routing 
> > > tables of some network domain are in sync ? 
> > > Or at least, that the topological view, aquired by all nodes 
> > > of some network domain complies with the physical reality ? 
> > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > be better than the religious fights on OAM-MPLS. 
> > > 
> > > Heinrich Hummel 
> > > Siemens AG 
> > > 
> > > 
> > > 
> > >  On Tue, 4 Jun 2002, Jing Shen wrote: 
> > > 
> > > > Given a network with several routers, is there any 
> > > tool/application to 
> > > > detect loops in the routing tables? 
> > > 
> > > As far as I know, there isn't. 
> > > Also if the loop is due to a misconfiguration which does 
> not show up 
> > > in BGP updates, then it's harder to detect. 
> > > 
> > > -Sue 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Mon Jul  8 13:42:10 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA27371
	for <routing-discussion-archive@odin.ietf.org>; Mon, 8 Jul 2002 13:42:09 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA07397;
	Mon, 8 Jul 2002 13:35:52 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA07368
	for <routing-discussion@optimus.ietf.org>; Mon, 8 Jul 2002 13:35:50 -0400 (EDT)
Received: from zcars04e.ca.nortel.com (zcars04e.nortelnetworks.com [47.129.242.56])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA26858
	for <routing-discussion@ietf.org>; Mon, 8 Jul 2002 13:34:58 -0400 (EDT)
Received: from zcard015.ca.nortel.com (zcard015.ca.nortel.com [47.129.30.7])
	by zcars04e.ca.nortel.com (Switch-2.2.0/Switch-2.2.0) with ESMTP id g68HZF903689;
	Mon, 8 Jul 2002 13:35:15 -0400 (EDT)
Received: by zcard015.ca.nortel.com with Internet Mail Service (5.5.2653.19)
	id <NYVCBBR3>; Mon, 8 Jul 2002 13:35:14 -0400
Message-ID: <9FBD322B7824D511B36900508BF93C9C04874ECF@zcard031.ca.nortel.com>
From: "David Allan" <dallan@nortelnetworks.com>
To: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org, "'jkr@visi.com'" <jkr@visi.com>,
        "'amir@cwnt.com'" <amir@cwnt.com>
Subject: RE: Loop Detection
Date: Mon, 8 Jul 2002 13:35:07 -0400 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: multipart/alternative;
	boundary="----_=_NextPart_001_01C226A5.CD85C2F0"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_001_01C226A5.CD85C2F0
Content-Type: text/plain;
	charset="iso-8859-1"

Heinrich:

I'd be concerned with scalability, as the derivative by-products are usually
context sensitive, e.g. unique to a node because of its position in the
topology, so whereas scaling a comparison of commonly flooded information is
a very tractable problem (simple exchange of digests), comparing node-unique
derivatives is not...

rgds
Dave

> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Monday, July 08, 2002 11:15 AM
> To: routing-discussion@ietf.org; 'jkr@visi.com'; 
> 'amir@cwnt.com'; Allan,
> David [CAR:NS00:EXCH]
> Subject: Re: Loop Detection
> 
> 
> It is not about TTL (which is to control a particular 
> IP-packet on its way to the destination).
> Its about a  sanity check w.r.t. the distributed view of the 
> intra-domain network topology. Whether or not IP flows are
> on the go is something else (they may do TTL). 
> 
> This sanity check may be simple and fast. And if you want, 
> there may be different types of sanity checks:
> some which controls just the topology without reachability 
> info,and  without link parameters. It will be the fastest
> (microseconds). Next there may be slower but more comprising 
> sanity checks. At last there may even be checks on
> proper derivatives (Routing Tables, NHLFEs,...)
> Heinrich
> 
> 
> 
> 
> Title: RE: Loop Detection
> Issue is it only tells you ***somthing*** is broken. 
> 
> Similarly simply ensuring that the routing databases are 
> synchronized does not perform a sanity check on the local 
> derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that 
> implements them.
> 
> cheers 
> Dave 
> 
> > -----Original Message----- 
> > From: John Renwick [mailto:jkr@visi.com] 
> > Sent: Wednesday, July 03, 2002 6:59 PM 
> > To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org 
> > Subject: RE: Loop Detection 
> > 
> > 
> > Isn't TTL a very direct and immediate form of loop detection?  
> > If you had a network tool that would illuminate the routers that 
> > are counting a lot of packets dropped because of TTL expiration, 
> > it could be showing you points on a routing loop. 
> > 
> > -- 
> > John Renwick                          Agere Systems 
> > Software Technical Manager            1230 Northland Drive 
> > Phone: (651) 675-2986                 Mendota Heights, MN 55120 
> > FAX:   (707) 897-7928                 Cell:  (612) 669-9397 
> > mailto:jrenwick@agere.com             http://www.agere.com 
> > 
> > > -----Original Message----- 
> > > From: routing-discussion-admin@ietf.org 
> > > [mailto:routing-discussion-admin@ietf.org]On Behalf Of 
> Amir Hermelin 
> > > Sent: Wednesday, July 03, 2002 5:12 PM 
> > > To: Hummel Heinrich; routing-discussion@ietf.org 
> > > Subject: RE: Loop Detection 
> > > 
> > > 
> > > my 2 cents: 
> > > * Yes, such a tool would be very nice. Although, a small portion 
> > > of our salaries is due to the fact that routing loops are 
> > so scary ;-) 
> > > * The approaches would be different for types of protocols. In 
> > > IGPs, assuming correct implementations, the protocol takes care 
> > > of loop possibilities as long as the lspdb is in sync. When its 
> > > not due to transient states, a tool wouldn't help since you'd 
> > > have to run it constantly for every router's view. 
> > > * I agree that detecting BGP mis-configuration is hard. As for 
> > > MPLS, there are methods that are more susceptible to loops than 
> > > others (e.g. one implementation of TE-shortcuts can lead to loops 
> > > even in a seemingly correct configuration, due to metric values). 
> > > * If this stuff moves forward, it might even mandate a whole new 
> > > workgroup. 
> > > 
> > > -- 
> > > Amir Hermelin                  <mailto:amir@cwnt.com> 
> > > Charlotte's Web Networks Inc.   <http://www.cwnt.com> 
> > > 
> > > 
> > > > -----Original Message----- 
> > > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de] 
> > > > Sent: Tuesday, July 02, 2002 11:17 AM 
> > > > To: routing-discussion@ietf.org 
> > > > Subject: Re: Loop Detection 
> > > > 
> > > > 
> > > > Wouldn't it be a fine goal to standardize a 
> > tool/mechanism/protocol 
> > > > which double-checks, within milliseconds, that all routing 
> > > > tables of some network domain are in sync ? 
> > > > Or at least, that the topological view, aquired by all nodes 
> > > > of some network domain complies with the physical reality ? 
> > > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > > be better than the religious fights on OAM-MPLS. 
> > > > 
> > > > Heinrich Hummel 
> > > > Siemens AG 
> > > > 
> > > > 
> > > > 
> > > >  On Tue, 4 Jun 2002, Jing Shen wrote: 
> > > > 
> > > > > Given a network with several routers, is there any 
> > > > tool/application to 
> > > > > detect loops in the routing tables? 
> > > > 
> > > > As far as I know, there isn't. 
> > > > Also if the loop is due to a misconfiguration which does 
> > not show up 
> > > > in BGP updates, then it's harder to detect. 
> > > > 
> > > > -Sue 
> 

------_=_NextPart_001_01C226A5.CD85C2F0
Content-Type: text/html;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Diso-8859-1">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
5.5.2655.35">
<TITLE>RE: Loop Detection</TITLE>
</HEAD>
<BODY>

<P><FONT SIZE=3D2>Heinrich:</FONT>
</P>

<P><FONT SIZE=3D2>I'd be concerned with scalability, as the derivative =
by-products are usually context sensitive, e.g. unique to a node =
because of its position in the topology, so whereas scaling a =
comparison of commonly flooded information is a very tractable problem =
(simple exchange of digests), comparing node-unique derivatives is =
not...</FONT></P>

<P><FONT SIZE=3D2>rgds</FONT>
<BR><FONT SIZE=3D2>Dave</FONT>
</P>

<P><FONT SIZE=3D2>&gt; -----Original Message-----</FONT>
<BR><FONT SIZE=3D2>&gt; From: Hummel Heinrich [<A =
HREF=3D"mailto:Heinrich.Hummel@icn.siemens.de">mailto:Heinrich.Hummel@ic=
n.siemens.de</A>]</FONT>
<BR><FONT SIZE=3D2>&gt; Sent: Monday, July 08, 2002 11:15 AM</FONT>
<BR><FONT SIZE=3D2>&gt; To: routing-discussion@ietf.org; =
'jkr@visi.com'; </FONT>
<BR><FONT SIZE=3D2>&gt; 'amir@cwnt.com'; Allan,</FONT>
<BR><FONT SIZE=3D2>&gt; David [CAR:NS00:EXCH]</FONT>
<BR><FONT SIZE=3D2>&gt; Subject: Re: Loop Detection</FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; It is not about TTL (which is to control a =
particular </FONT>
<BR><FONT SIZE=3D2>&gt; IP-packet on its way to the =
destination).</FONT>
<BR><FONT SIZE=3D2>&gt; Its about a&nbsp; sanity check w.r.t. the =
distributed view of the </FONT>
<BR><FONT SIZE=3D2>&gt; intra-domain network topology. Whether or not =
IP flows are</FONT>
<BR><FONT SIZE=3D2>&gt; on the go is something else (they may do TTL). =
</FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; This sanity check may be simple and fast. And =
if you want, </FONT>
<BR><FONT SIZE=3D2>&gt; there may be different types of sanity =
checks:</FONT>
<BR><FONT SIZE=3D2>&gt; some which controls just the topology without =
reachability </FONT>
<BR><FONT SIZE=3D2>&gt; info,and&nbsp; without link parameters. It will =
be the fastest</FONT>
<BR><FONT SIZE=3D2>&gt; (microseconds). Next there may be slower but =
more comprising </FONT>
<BR><FONT SIZE=3D2>&gt; sanity checks. At last there may even be checks =
on</FONT>
<BR><FONT SIZE=3D2>&gt; proper derivatives (Routing Tables, =
NHLFEs,...)</FONT>
<BR><FONT SIZE=3D2>&gt; Heinrich</FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; Title: RE: Loop Detection</FONT>
<BR><FONT SIZE=3D2>&gt; Issue is it only tells you ***somthing*** is =
broken. </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; Similarly simply ensuring that the routing =
databases are </FONT>
<BR><FONT SIZE=3D2>&gt; synchronized does not perform a sanity check on =
the local </FONT>
<BR><FONT SIZE=3D2>&gt; derivative by-products (FIBs, FTN, NHLFE etc.) =
or the hw that </FONT>
<BR><FONT SIZE=3D2>&gt; implements them.</FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; cheers </FONT>
<BR><FONT SIZE=3D2>&gt; Dave </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; -----Original Message----- </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; From: John Renwick [<A =
HREF=3D"mailto:jkr@visi.com">mailto:jkr@visi.com</A>] </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Sent: Wednesday, July 03, 2002 6:59 PM =
</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; To: Amir Hermelin; Hummel Heinrich; =
routing-discussion@ietf.org </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Subject: RE: Loop Detection </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Isn't TTL a very direct and immediate form =
of loop detection?&nbsp; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; If you had a network tool that would =
illuminate the routers that </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; are counting a lot of packets dropped =
because of TTL expiration, </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; it could be showing you points on a =
routing loop. </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; -- </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; John =
Renwick&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
;&nbsp;&nbsp; Agere Systems </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Software Technical =
Manager&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
; 1230 Northland Drive </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; Phone: (651) =
675-2986&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Mendota Heights, MN 55120 </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; FAX:&nbsp;&nbsp; (707) =
897-7928&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Cell:&nbsp; (612) 669-9397 </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; <A =
HREF=3D"mailto:jrenwick@agere.com">mailto:jrenwick@agere.com</A>&nbsp;&n=
bsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; <A =
HREF=3D"http://www.agere.com" =
TARGET=3D"_blank">http://www.agere.com</A> </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; -----Original Message----- </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; From: =
routing-discussion-admin@ietf.org </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; [<A =
HREF=3D"mailto:routing-discussion-admin@ietf.org">mailto:routing-discuss=
ion-admin@ietf.org</A>]On Behalf Of </FONT>
<BR><FONT SIZE=3D2>&gt; Amir Hermelin </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Sent: Wednesday, July 03, 2002 5:12 =
PM </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; To: Hummel Heinrich; =
routing-discussion@ietf.org </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Subject: RE: Loop Detection </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; my 2 cents: </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; * Yes, such a tool would be very =
nice. Although, a small portion </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; of our salaries is due to the fact =
that routing loops are </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; so scary ;-) </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; * The approaches would be different =
for types of protocols. In </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; IGPs, assuming correct =
implementations, the protocol takes care </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; of loop possibilities as long as the =
lspdb is in sync. When its </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; not due to transient states, a tool =
wouldn't help since you'd </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; have to run it constantly for every =
router's view. </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; * I agree that detecting BGP =
mis-configuration is hard. As for </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; MPLS, there are methods that are more =
susceptible to loops than </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; others (e.g. one implementation of =
TE-shortcuts can lead to loops </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; even in a seemingly correct =
configuration, due to metric values). </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; * If this stuff moves forward, it =
might even mandate a whole new </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; workgroup. </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; -- </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Amir =
Hermelin&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbs=
p;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; &lt;<A =
HREF=3D"mailto:amir@cwnt.com">mailto:amir@cwnt.com</A>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; Charlotte's Web Networks =
Inc.&nbsp;&nbsp; &lt;<A HREF=3D"http://www.cwnt.com" =
TARGET=3D"_blank">http://www.cwnt.com</A>&gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; -----Original Message----- =
</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; From: Hummel Heinrich [<A =
HREF=3D"mailto:Heinrich.Hummel@icn.siemens.de">mailto:Heinrich.Hummel@ic=
n.siemens.de</A>] </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Sent: Tuesday, July 02, 2002 =
11:17 AM </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; To: routing-discussion@ietf.org =
</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Subject: Re: Loop Detection =
</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Wouldn't it be a fine goal to =
standardize a </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; tool/mechanism/protocol </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; which double-checks, within =
milliseconds, that all routing </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; tables of some network domain =
are in sync ? </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Or at least, that the =
topological view, aquired by all nodes </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; of some network domain complies =
with the physical reality ? </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Wouldn't this be a powerful =
OAM-application ? IMHO:it would </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; be better than the religious =
fights on OAM-MPLS. </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Heinrich Hummel </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Siemens AG </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt;&nbsp; On Tue, 4 Jun 2002, Jing =
Shen wrote: </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; &gt; Given a network with =
several routers, is there any </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; tool/application to </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; &gt; detect loops in the routing =
tables? </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; As far as I know, there isn't. =
</FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; Also if the loop is due to a =
misconfiguration which does </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; not show up </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; in BGP updates, then it's harder =
to detect. </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; </FONT>
<BR><FONT SIZE=3D2>&gt; &gt; &gt; &gt; -Sue </FONT>
<BR><FONT SIZE=3D2>&gt; </FONT>
</P>

</BODY>
</HTML>
------_=_NextPart_001_01C226A5.CD85C2F0--

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Tue Jul  9 17:08:07 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA17555
	for <routing-discussion-archive@odin.ietf.org>; Tue, 9 Jul 2002 17:08:07 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id QAA06991;
	Tue, 9 Jul 2002 16:47:31 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id QAA06923
	for <routing-discussion@optimus.ietf.org>; Tue, 9 Jul 2002 16:47:27 -0400 (EDT)
Received: from psg.com (exim@psg.com [147.28.0.62])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA16938;
	Tue, 9 Jul 2002 16:46:31 -0400 (EDT)
Received: from 12-234-206-11.client.attbi.com ([12.234.206.11] helo=AGNS3K11)
	by psg.com with esmtp (Exim 3.36 #1)
	id 17S1sX-0005RO-00; Tue, 09 Jul 2002 13:46:33 -0700
Date: Tue, 9 Jul 2002 13:46:03 -0700
From: Alex Zinin <zinin@psg.com>
X-Mailer: The Bat! (v1.51) Personal
Reply-To: Alex Zinin <zinin@psg.com>
X-Priority: 3 (Normal)
Message-ID: <10875018761.20020709134603@psg.com>
To: rpsec@ietf.org
CC: routing-discussion@ietf.org
Subject: Fwd: WG Action: Routing Protocol Security Requirements (rpsec)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Folks,

 In case someone missed the announcement, RPSEC has been
 approved as a WG.

 Regarding the chairmanship--we're still in the process
 here. Bill kindly agreed to help as the co-chair until
 we're there.

 See you in Yokohama.

Alex


This is a forwarded message
From: Steve Coya <scoya@cnri.reston.va.us>
To: 
Cc: 
Date: Tuesday, July 02, 2002, 6:12:10 AM
Subject: WG Action: Routing Protocol Security Requirements (rpsec)

===8<==============Original message text===============

A new IETF working group has been established in the Routing Area.
Please contact the Area Directors or WG CHairs for additional
information.

Routing Protocol Security Requirements (rpsec)
----------------------------------------------

 Current Status: Active Working Group

 Chair(s):
     Bill Fenner  <fenner@research.att.com>
     Russ White  <riw@cisco.com>

 Routing Area Director(s):
     Bill Fenner  <fenner@research.att.com>
     Alex Zinin  <zinin@psg.com>

 Routing Area Advisor:
     Alex Zinin  <zinin@psg.com>

 Mailing Lists: 
     General Discussion:rpsec@ietf.org
     To Subscribe:      rpsec-request@ietf.org
         In Body:       (un)subscribe
     Archive:           http://www1.ietf.org/mail-archive/working-groups/rpsec

Description of Working Group:

The lack of a common set of security requirements and methods for
routing protocols has resulted in a wide variety of security mechanisms
for individual routing protocols. Ongoing work on requirements for the
next generation routing system and future work on the actual mechanisms
for it will require well documented routing security requirements.

The products of this working group will be used by routing protocol
designers to ensure adequate coverage of security in the future,
including well known and possible threats.

The scope of work is limited to router-to-router protocols only for
both unicast and multicast systems, and does NOT include host-to-router
protocol such as IGMP, ICMP, ARP, or ND. It is also a non-goal at this
point to produce new or change the current security mechanisms in the
existing routing protocols.

The RPSEC working group is charged with the following tasks:

- Document threat models for routing systems

- Document security requirements for routing systems

- Provide a common area for discussion between security and routing
  experts on the topic of securing the routing system

Possible Future Work

- Evaluate and document existing and proposed routing security
  mechanisms with respect to established RPSEC requirements

- Recommend mechanism(s)

 Goals and Milestones:

   JUL 02       Submit initial I-D (or set of I-Ds) which details the 
                threats to routing systems. 

   OCT 02       Submit initial I-D (or set of I-Ds) which outlines security 
                requirements for routing systems. 

   DEC 02       Submit I-Ds documenting threats to routing systems for 
                publication as Informational RFC. 

   MAR 03       Submit the I-D documenting security requirements to routing 
                systems for publication as Informational RFC. 

   MAR 03       Evaluate progress, recharter with new goals (see possible 
                future work above) or shutdown. 


===8<===========End of original message text===========



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Tue Jul  9 22:28:18 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA24816
	for <routing-discussion-archive@odin.ietf.org>; Tue, 9 Jul 2002 22:28:18 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA27063;
	Tue, 9 Jul 2002 22:19:52 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA27034
	for <routing-discussion@optimus.ietf.org>; Tue, 9 Jul 2002 22:19:50 -0400 (EDT)
Received: from hoemail1.firewall.lucent.com (hoemail1.lucent.com [192.11.226.161])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA24605
	for <routing-discussion@ietf.org>; Tue, 9 Jul 2002 22:18:57 -0400 (EDT)
Received: from ci1093exch001p.wins.lucent.com (h135-252-12-239.lucent.com [135.252.12.239])
	by hoemail1.firewall.lucent.com (Switch-2.2.2/Switch-2.2.0) with ESMTP id g6A2JIO29954
	for <routing-discussion@ietf.org>; Tue, 9 Jul 2002 22:19:18 -0400 (EDT)
Received: by CI1093EXCH001P with Internet Mail Service (5.5.2653.19)
	id <NJNKMHA7>; Wed, 10 Jul 2002 10:19:17 +0800
Message-ID: <31C0F08B0D18D511ACC800508BAE7B4702B9D013@CI0026EXCH001U>
From: "Li, Ke Qin (Peter)" <keqinli@lucent.com>
To: "'Hummel Heinrich'" <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: RE: Loop Detection
Date: Wed, 10 Jul 2002 10:19:11 +0800
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Yes, it is not about TTL.

Is there any existing work about the sanity check?

And would you please give more detailed explaination about the 3 kinds of
sanity check?

Keqin Li

> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Monday, July 08, 2002 11:15 PM
> To: routing-discussion@ietf.org; 'jkr@visi.com'; 
> 'amir@cwnt.com'; 'dallan@nortelnetworks.com'
> Subject: Re: Loop Detection
> 
> 
> It is not about TTL (which is to control a particular 
> IP-packet on its way to the destination).
> Its about a  sanity check w.r.t. the distributed view of the 
> intra-domain network topology. Whether or not IP flows are
> on the go is something else (they may do TTL). 
> 
> This sanity check may be simple and fast. And if you want, 
> there may be different types of sanity checks:
> some which controls just the topology without reachability 
> info,and  without link parameters. It will be the fastest
> (microseconds). Next there may be slower but more comprising 
> sanity checks. At last there may even be checks on
> proper derivatives (Routing Tables, NHLFEs,...)
> Heinrich
> 
> 
> 
> 
> Title: RE: Loop Detection
> Issue is it only tells you ***somthing*** is broken. 
> 
> Similarly simply ensuring that the routing databases are 
> synchronized does not perform a sanity check on the local 
> derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that 
> implements them.
> 
> cheers 
> Dave 
> 
> > -----Original Message----- 
> > From: John Renwick [mailto:jkr@visi.com] 
> > Sent: Wednesday, July 03, 2002 6:59 PM 
> > To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org 
> > Subject: RE: Loop Detection 
> > 
> > 
> > Isn't TTL a very direct and immediate form of loop detection?  
> > If you had a network tool that would illuminate the routers that 
> > are counting a lot of packets dropped because of TTL expiration, 
> > it could be showing you points on a routing loop. 
> > 
> > -- 
> > John Renwick                          Agere Systems 
> > Software Technical Manager            1230 Northland Drive 
> > Phone: (651) 675-2986                 Mendota Heights, MN 55120 
> > FAX:   (707) 897-7928                 Cell:  (612) 669-9397 
> > mailto:jrenwick@agere.com             http://www.agere.com 
> > 
> > > -----Original Message----- 
> > > From: routing-discussion-admin@ietf.org 
> > > [mailto:routing-discussion-admin@ietf.org]On Behalf Of 
> Amir Hermelin 
> > > Sent: Wednesday, July 03, 2002 5:12 PM 
> > > To: Hummel Heinrich; routing-discussion@ietf.org 
> > > Subject: RE: Loop Detection 
> > > 
> > > 
> > > my 2 cents: 
> > > * Yes, such a tool would be very nice. Although, a small portion 
> > > of our salaries is due to the fact that routing loops are 
> > so scary ;-) 
> > > * The approaches would be different for types of protocols. In 
> > > IGPs, assuming correct implementations, the protocol takes care 
> > > of loop possibilities as long as the lspdb is in sync. When its 
> > > not due to transient states, a tool wouldn't help since you'd 
> > > have to run it constantly for every router's view. 
> > > * I agree that detecting BGP mis-configuration is hard. As for 
> > > MPLS, there are methods that are more susceptible to loops than 
> > > others (e.g. one implementation of TE-shortcuts can lead to loops 
> > > even in a seemingly correct configuration, due to metric values). 
> > > * If this stuff moves forward, it might even mandate a whole new 
> > > workgroup. 
> > > 
> > > -- 
> > > Amir Hermelin                  <mailto:amir@cwnt.com> 
> > > Charlotte's Web Networks Inc.   <http://www.cwnt.com> 
> > > 
> > > 
> > > > -----Original Message----- 
> > > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de] 
> > > > Sent: Tuesday, July 02, 2002 11:17 AM 
> > > > To: routing-discussion@ietf.org 
> > > > Subject: Re: Loop Detection 
> > > > 
> > > > 
> > > > Wouldn't it be a fine goal to standardize a 
> > tool/mechanism/protocol 
> > > > which double-checks, within milliseconds, that all routing 
> > > > tables of some network domain are in sync ? 
> > > > Or at least, that the topological view, aquired by all nodes 
> > > > of some network domain complies with the physical reality ? 
> > > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > > be better than the religious fights on OAM-MPLS. 
> > > > 
> > > > Heinrich Hummel 
> > > > Siemens AG 
> > > > 
> > > > 
> > > > 
> > > >  On Tue, 4 Jun 2002, Jing Shen wrote: 
> > > > 
> > > > > Given a network with several routers, is there any 
> > > > tool/application to 
> > > > > detect loops in the routing tables? 
> > > > 
> > > > As far as I know, there isn't. 
> > > > Also if the loop is due to a misconfiguration which does 
> > not show up 
> > > > in BGP updates, then it's harder to detect. 
> > > > 
> > > > -Sue 
> 
> _______________________________________________
> routing-discussion mailing list
> routing-discussion@ietf.org
> https://www1.ietf.org/mailman/listinfo/routing-discussion
> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Wed Jul 10 03:11:24 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA23431
	for <routing-discussion-archive@odin.ietf.org>; Wed, 10 Jul 2002 03:11:24 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id CAA01581;
	Wed, 10 Jul 2002 02:59:26 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id CAA01550
	for <routing-discussion@optimus.ietf.org>; Wed, 10 Jul 2002 02:59:24 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id CAA23118
	for <routing-discussion@ietf.org>; Wed, 10 Jul 2002 02:58:30 -0400 (EDT)
Received: from blues.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.227])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id IAA01451;
	Wed, 10 Jul 2002 08:59:00 +0200 (MET DST)
Received: from mchh273e.demchh201e.icn.siemens.de ([139.21.200.83])
	by blues.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id IAA12727;
	Wed, 10 Jul 2002 08:59:15 +0200 (MET DST)
Received: by MCHH273E with Internet Mail Service (5.5.2653.19)
	id <MY3DX9PB>; Wed, 10 Jul 2002 08:59:12 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B8FF@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Li, Ke Qin (Peter)'" <keqinli@lucent.com>,
        Hummel Heinrich
	 <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: AW: Loop Detection
Date: Wed, 10 Jul 2002 08:59:02 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

See my answers inserted.
Heinrich
 


Yes, it is not about TTL.

Is there any existing work about the sanity check?

HH=> I haven't heard of any work being done in the IETF. And I have just asked the question
whether there were interest on doing such work.

And would you please give more detailed explaination about the 3 kinds of
sanity check?

HH=> Sure, but before, I would suggest to agree on a different subject line than "Loop Detection"
because it is really not about TTL. Instead it is about checking whether all nodes have the identical view of the network.
Routing tables don't have loops but may be inconsistent.
E.g. "OAM: Network view sanity check" might be a more appropriate subject line.

Also,the less you check, the faster and simpler is the sanity check. I first of all thought about checking the
topological view only/mainly.





Keqin Li

> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Monday, July 08, 2002 11:15 PM
> To: routing-discussion@ietf.org; 'jkr@visi.com'; 
> 'amir@cwnt.com'; 'dallan@nortelnetworks.com'
> Subject: Re: Loop Detection
> 
> 
> It is not about TTL (which is to control a particular 
> IP-packet on its way to the destination).
> Its about a  sanity check w.r.t. the distributed view of the 
> intra-domain network topology. Whether or not IP flows are
> on the go is something else (they may do TTL). 
> 
> This sanity check may be simple and fast. And if you want, 
> there may be different types of sanity checks:
> some which controls just the topology without reachability 
> info,and  without link parameters. It will be the fastest
> (microseconds). Next there may be slower but more comprising 
> sanity checks. At last there may even be checks on
> proper derivatives (Routing Tables, NHLFEs,...)
> Heinrich
> 
> 
> 
> 
> Title: RE: Loop Detection
> Issue is it only tells you ***somthing*** is broken. 
> 
> Similarly simply ensuring that the routing databases are 
> synchronized does not perform a sanity check on the local 
> derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that 
> implements them.
> 
> cheers 
> Dave 
> 
> > -----Original Message----- 
> > From: John Renwick [mailto:jkr@visi.com] 
> > Sent: Wednesday, July 03, 2002 6:59 PM 
> > To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org 
> > Subject: RE: Loop Detection 
> > 
> > 
> > Isn't TTL a very direct and immediate form of loop detection?  
> > If you had a network tool that would illuminate the routers that 
> > are counting a lot of packets dropped because of TTL expiration, 
> > it could be showing you points on a routing loop. 
> > 
> > -- 
> > John Renwick                          Agere Systems 
> > Software Technical Manager            1230 Northland Drive 
> > Phone: (651) 675-2986                 Mendota Heights, MN 55120 
> > FAX:   (707) 897-7928                 Cell:  (612) 669-9397 
> > mailto:jrenwick@agere.com             http://www.agere.com 
> > 
> > > -----Original Message----- 
> > > From: routing-discussion-admin@ietf.org 
> > > [mailto:routing-discussion-admin@ietf.org]On Behalf Of 
> Amir Hermelin 
> > > Sent: Wednesday, July 03, 2002 5:12 PM 
> > > To: Hummel Heinrich; routing-discussion@ietf.org 
> > > Subject: RE: Loop Detection 
> > > 
> > > 
> > > my 2 cents: 
> > > * Yes, such a tool would be very nice. Although, a small portion 
> > > of our salaries is due to the fact that routing loops are 
> > so scary ;-) 
> > > * The approaches would be different for types of protocols. In 
> > > IGPs, assuming correct implementations, the protocol takes care 
> > > of loop possibilities as long as the lspdb is in sync. When its 
> > > not due to transient states, a tool wouldn't help since you'd 
> > > have to run it constantly for every router's view. 
> > > * I agree that detecting BGP mis-configuration is hard. As for 
> > > MPLS, there are methods that are more susceptible to loops than 
> > > others (e.g. one implementation of TE-shortcuts can lead to loops 
> > > even in a seemingly correct configuration, due to metric values). 
> > > * If this stuff moves forward, it might even mandate a whole new 
> > > workgroup. 
> > > 
> > > -- 
> > > Amir Hermelin                  <mailto:amir@cwnt.com> 
> > > Charlotte's Web Networks Inc.   <http://www.cwnt.com> 
> > > 
> > > 
> > > > -----Original Message----- 
> > > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de] 
> > > > Sent: Tuesday, July 02, 2002 11:17 AM 
> > > > To: routing-discussion@ietf.org 
> > > > Subject: Re: Loop Detection 
> > > > 
> > > > 
> > > > Wouldn't it be a fine goal to standardize a 
> > tool/mechanism/protocol 
> > > > which double-checks, within milliseconds, that all routing 
> > > > tables of some network domain are in sync ? 
> > > > Or at least, that the topological view, aquired by all nodes 
> > > > of some network domain complies with the physical reality ? 
> > > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > > be better than the religious fights on OAM-MPLS. 
> > > > 
> > > > Heinrich Hummel 
> > > > Siemens AG 
> > > > 
> > > > 
> > > > 
> > > >  On Tue, 4 Jun 2002, Jing Shen wrote: 
> > > > 
> > > > > Given a network with several routers, is there any 
> > > > tool/application to 
> > > > > detect loops in the routing tables? 
> > > > 
> > > > As far as I know, there isn't. 
> > > > Also if the loop is due to a misconfiguration which does 
> > not show up 
> > > > in BGP updates, then it's harder to detect. 
> > > > 
> > > > -Sue 
> 
> _______________________________________________
> routing-discussion mailing list
> routing-discussion@ietf.org
> https://www1.ietf.org/mailman/listinfo/routing-discussion
> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Wed Jul 10 22:26:39 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA27387
	for <routing-discussion-archive@odin.ietf.org>; Wed, 10 Jul 2002 22:26:39 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA24851;
	Wed, 10 Jul 2002 22:20:01 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA24825
	for <routing-discussion@optimus.ietf.org>; Wed, 10 Jul 2002 22:19:59 -0400 (EDT)
Received: from auemail1.firewall.lucent.com (auemail1.lucent.com [192.11.223.161])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA27155
	for <routing-discussion@ietf.org>; Wed, 10 Jul 2002 22:19:05 -0400 (EDT)
Received: from ci1093exch001p.wins.lucent.com (h135-252-12-239.lucent.com [135.252.12.239])
	by auemail1.firewall.lucent.com (Switch-2.2.2/Switch-2.2.0) with ESMTP id g6B2JQq00818
	for <routing-discussion@ietf.org>; Wed, 10 Jul 2002 22:19:26 -0400 (EDT)
Received: by CI1093EXCH001P with Internet Mail Service (5.5.2653.19)
	id <NJNKM5R9>; Thu, 11 Jul 2002 10:19:12 +0800
Message-ID: <31C0F08B0D18D511ACC800508BAE7B4702B9D014@CI0026EXCH001U>
From: "Li, Ke Qin (Peter)" <keqinli@lucent.com>
To: "'Hummel Heinrich'" <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: OAM: Network view sanity check
Date: Thu, 11 Jul 2002 10:18:46 +0800
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

As a new comer, I don't know the meaning of OAM. Would you please explain it
for me? Thank you.

The next question is how to define "nodes' view of network". In my opinion,
the sanity check technique should be routing porotocol independent. Thus,
only the routing table (even the forwarding table) can be checked. And the
routing information database which describes the routing protocol's view of
network topology is not accessible. Then, how to define "nodes' view of
network"?

Keqin Li

> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Wednesday, July 10, 2002 2:59 PM
> To: 'Li, Ke Qin (Peter)'; Hummel Heinrich; routing-discussion@ietf.org
> Subject: AW: Loop Detection
> 
> 
> See my answers inserted.
> Heinrich
>  
> 
> 
> Yes, it is not about TTL.
> 
> Is there any existing work about the sanity check?
> 
> HH=> I haven't heard of any work being done in the IETF. And 
> I have just asked the question
> whether there were interest on doing such work.
> 
> And would you please give more detailed explaination about 
> the 3 kinds of
> sanity check?
> 
> HH=> Sure, but before, I would suggest to agree on a 
> different subject line than "Loop Detection"
> because it is really not about TTL. Instead it is about 
> checking whether all nodes have the identical view of the network.
> Routing tables don't have loops but may be inconsistent.
> E.g. "OAM: Network view sanity check" might be a more 
> appropriate subject line.
> 
> Also,the less you check, the faster and simpler is the sanity 
> check. I first of all thought about checking the
> topological view only/mainly.
> 
> 
> 
> 
> 
> Keqin Li
> 
> > -----Original Message-----
> > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> > Sent: Monday, July 08, 2002 11:15 PM
> > To: routing-discussion@ietf.org; 'jkr@visi.com'; 
> > 'amir@cwnt.com'; 'dallan@nortelnetworks.com'
> > Subject: Re: Loop Detection
> > 
> > 
> > It is not about TTL (which is to control a particular 
> > IP-packet on its way to the destination).
> > Its about a  sanity check w.r.t. the distributed view of the 
> > intra-domain network topology. Whether or not IP flows are
> > on the go is something else (they may do TTL). 
> > 
> > This sanity check may be simple and fast. And if you want, 
> > there may be different types of sanity checks:
> > some which controls just the topology without reachability 
> > info,and  without link parameters. It will be the fastest
> > (microseconds). Next there may be slower but more comprising 
> > sanity checks. At last there may even be checks on
> > proper derivatives (Routing Tables, NHLFEs,...)
> > Heinrich
> > 
> > 
> > 
> > 
> > Title: RE: Loop Detection
> > Issue is it only tells you ***somthing*** is broken. 
> > 
> > Similarly simply ensuring that the routing databases are 
> > synchronized does not perform a sanity check on the local 
> > derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that 
> > implements them.
> > 
> > cheers 
> > Dave 
> > 
> > > -----Original Message----- 
> > > From: John Renwick [mailto:jkr@visi.com] 
> > > Sent: Wednesday, July 03, 2002 6:59 PM 
> > > To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org 
> > > Subject: RE: Loop Detection 
> > > 
> > > 
> > > Isn't TTL a very direct and immediate form of loop detection?  
> > > If you had a network tool that would illuminate the routers that 
> > > are counting a lot of packets dropped because of TTL expiration, 
> > > it could be showing you points on a routing loop. 
> > > 
> > > -- 
> > > John Renwick                          Agere Systems 
> > > Software Technical Manager            1230 Northland Drive 
> > > Phone: (651) 675-2986                 Mendota Heights, MN 55120 
> > > FAX:   (707) 897-7928                 Cell:  (612) 669-9397 
> > > mailto:jrenwick@agere.com             http://www.agere.com 
> > > 
> > > > -----Original Message----- 
> > > > From: routing-discussion-admin@ietf.org 
> > > > [mailto:routing-discussion-admin@ietf.org]On Behalf Of 
> > Amir Hermelin 
> > > > Sent: Wednesday, July 03, 2002 5:12 PM 
> > > > To: Hummel Heinrich; routing-discussion@ietf.org 
> > > > Subject: RE: Loop Detection 
> > > > 
> > > > 
> > > > my 2 cents: 
> > > > * Yes, such a tool would be very nice. Although, a 
> small portion 
> > > > of our salaries is due to the fact that routing loops are 
> > > so scary ;-) 
> > > > * The approaches would be different for types of protocols. In 
> > > > IGPs, assuming correct implementations, the protocol takes care 
> > > > of loop possibilities as long as the lspdb is in sync. When its 
> > > > not due to transient states, a tool wouldn't help since you'd 
> > > > have to run it constantly for every router's view. 
> > > > * I agree that detecting BGP mis-configuration is hard. As for 
> > > > MPLS, there are methods that are more susceptible to loops than 
> > > > others (e.g. one implementation of TE-shortcuts can 
> lead to loops 
> > > > even in a seemingly correct configuration, due to 
> metric values). 
> > > > * If this stuff moves forward, it might even mandate a 
> whole new 
> > > > workgroup. 
> > > > 
> > > > -- 
> > > > Amir Hermelin                  <mailto:amir@cwnt.com> 
> > > > Charlotte's Web Networks Inc.   <http://www.cwnt.com> 
> > > > 
> > > > 
> > > > > -----Original Message----- 
> > > > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de] 
> > > > > Sent: Tuesday, July 02, 2002 11:17 AM 
> > > > > To: routing-discussion@ietf.org 
> > > > > Subject: Re: Loop Detection 
> > > > > 
> > > > > 
> > > > > Wouldn't it be a fine goal to standardize a 
> > > tool/mechanism/protocol 
> > > > > which double-checks, within milliseconds, that all routing 
> > > > > tables of some network domain are in sync ? 
> > > > > Or at least, that the topological view, aquired by all nodes 
> > > > > of some network domain complies with the physical reality ? 
> > > > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > > > be better than the religious fights on OAM-MPLS. 
> > > > > 
> > > > > Heinrich Hummel 
> > > > > Siemens AG 
> > > > > 
> > > > > 
> > > > > 
> > > > >  On Tue, 4 Jun 2002, Jing Shen wrote: 
> > > > > 
> > > > > > Given a network with several routers, is there any 
> > > > > tool/application to 
> > > > > > detect loops in the routing tables? 
> > > > > 
> > > > > As far as I know, there isn't. 
> > > > > Also if the loop is due to a misconfiguration which does 
> > > not show up 
> > > > > in BGP updates, then it's harder to detect. 
> > > > > 
> > > > > -Sue 
> > 
> > _______________________________________________
> > routing-discussion mailing list
> > routing-discussion@ietf.org
> > https://www1.ietf.org/mailman/listinfo/routing-discussion
> > 
> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Wed Jul 10 22:42:03 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA28194
	for <routing-discussion-archive@odin.ietf.org>; Wed, 10 Jul 2002 22:42:03 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA26058;
	Wed, 10 Jul 2002 22:34:21 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id VAA23536
	for <routing-discussion@optimus.ietf.org>; Wed, 10 Jul 2002 21:34:37 -0400 (EDT)
Received: from mail.cad.zju.edu.cn (cad.zju.edu.cn [210.32.131.2])
	by ietf.org (8.9.1a/8.9.1a) with SMTP id VAA25783
	for <routing-discussion@ietf.org>; Wed, 10 Jul 2002 21:33:36 -0400 (EDT)
Received: (qmail 7592 invoked from network); 11 Jul 2002 01:39:16 -0000
Received: from unknown (HELO cad.zju.edu.cn) (210.32.131.97)
  by 210.32.131.2 with SMTP; 11 Jul 2002 01:39:16 -0000
Message-ID: <3D2CE08C.60C13F19@cad.zju.edu.cn>
Date: Thu, 11 Jul 2002 09:34:04 +0800
From: Jing Shen <jshen@cad.zju.edu.cn>
Reply-To: jshen@cad.zju.edu.cn
Organization: State Key Lab of CAD&CG
X-Mailer: Mozilla 4.79 [en] (X11; U; SunOS 5.6 sun4u)
X-Accept-Language: en
MIME-Version: 1.0
To: Amir Hermelin <amir@cwnt.com>
CC: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: Re: Loop Detection
References: <C7DF4400240AFB4095D98C7C6EC2A34A73012B@bart.cwnt.com>
Content-Type: text/plain; charset=gb2312
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Why do you think MPLS is more susceptible to loop than other technology? 
As I think, LSP is set up by signalling along the path computed by edge
router or
following the items of routing tables. That is, if we just set up 
LSPs according to routing table of IGP or EGP, the occurance of loop
will be
the same level of  traditional routing. 

To detect loop in network, I think it's nearly impossible for a node to
detect
loops on some directions if it could not track routing tables at real
time. But,
if CR-LDP is used to set up transmission tunnel in networks, the problem
may 
be overcomed at the stage of setup.




Amir Hermelin wrote:
> 
> my 2 cents:
> * Yes, such a tool would be very nice. Although, a small portion of our salaries is due to the fact that routing loops are so scary ;-)
> * The approaches would be different for types of protocols. In IGPs, assuming correct implementations, the protocol takes care of loop possibilities as long as the lspdb is in sync. When its not due to transient states, a tool wouldn't help since you'd have to run it constantly for every router's view.
> * I agree that detecting BGP mis-configuration is hard. As for MPLS, there are methods that are more susceptible to loops than others (e.g. one implementation of TE-shortcuts can lead to loops even in a seemingly correct configuration, due to metric values).
> * If this stuff moves forward, it might even mandate a whole new workgroup.
> 
> --
> Amir Hermelin                  <mailto:amir@cwnt.com>
> Charlotte's Web Networks Inc.   <http://www.cwnt.com>
> 

-- 
Jing Shen

Email: jshen@cad.zju.edu.cn

**********************************************************************
* The SunShine of life is made up of very little beams which is      *
*  bright all the time                                               *
**********************************************************************


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Thu Jul 11 03:33:34 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA11901
	for <routing-discussion-archive@odin.ietf.org>; Thu, 11 Jul 2002 03:33:34 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id DAA17421;
	Thu, 11 Jul 2002 03:27:41 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id DAA17364
	for <routing-discussion@optimus.ietf.org>; Thu, 11 Jul 2002 03:27:38 -0400 (EDT)
Received: from gorilla.mchh.siemens.de (gorilla.mchh.siemens.de [194.138.158.18])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA11741
	for <routing-discussion@ietf.org>; Thu, 11 Jul 2002 03:26:43 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by gorilla.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id JAA02408;
	Thu, 11 Jul 2002 09:27:32 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id JAA29377;
	Thu, 11 Jul 2002 09:27:34 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2653.19)
	id <MYN1DA05>; Thu, 11 Jul 2002 09:27:34 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B90A@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Li, Ke Qin (Peter)'" <keqinli@lucent.com>, routing-discussion@ietf.org
Subject: AW: Network view sanity check
Date: Thu, 11 Jul 2002 09:27:33 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id DAA17365
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

OAM. I think its meaning is operation,administration,maintenance.

The sanity check should work no matter whether OSPF or IS-IS is used.
However it depends on the used routing protocol in the same way as CRC-bits depend on the transmitted, to be checked, text.
Heinrich


-----Ursprüngliche Nachricht-----
Von: Li, Ke Qin (Peter) [mailto:keqinli@lucent.com]
Gesendet: Donnerstag, 11. Juli 2002 04:19
An: 'Hummel Heinrich'; routing-discussion@ietf.org
Betreff: OAM: Network view sanity check


As a new comer, I don't know the meaning of OAM. Would you please explain it
for me? Thank you.

The next question is how to define "nodes' view of network". In my opinion,
the sanity check technique should be routing porotocol independent. Thus,
only the routing table (even the forwarding table) can be checked. And the
routing information database which describes the routing protocol's view of
network topology is not accessible. Then, how to define "nodes' view of
network"?

Keqin Li

> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Wednesday, July 10, 2002 2:59 PM
> To: 'Li, Ke Qin (Peter)'; Hummel Heinrich; routing-discussion@ietf.org
> Subject: AW: Loop Detection
> 
> 
> See my answers inserted.
> Heinrich
>  
> 
> 
> Yes, it is not about TTL.
> 
> Is there any existing work about the sanity check?
> 
> HH=> I haven't heard of any work being done in the IETF. And 
> I have just asked the question
> whether there were interest on doing such work.
> 
> And would you please give more detailed explaination about 
> the 3 kinds of
> sanity check?
> 
> HH=> Sure, but before, I would suggest to agree on a 
> different subject line than "Loop Detection"
> because it is really not about TTL. Instead it is about 
> checking whether all nodes have the identical view of the network.
> Routing tables don't have loops but may be inconsistent.
> E.g. "OAM: Network view sanity check" might be a more 
> appropriate subject line.
> 
> Also,the less you check, the faster and simpler is the sanity 
> check. I first of all thought about checking the
> topological view only/mainly.
> 
> 
> 
> 
> 
> Keqin Li
> 
> > -----Original Message-----
> > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> > Sent: Monday, July 08, 2002 11:15 PM
> > To: routing-discussion@ietf.org; 'jkr@visi.com'; 
> > 'amir@cwnt.com'; 'dallan@nortelnetworks.com'
> > Subject: Re: Loop Detection
> > 
> > 
> > It is not about TTL (which is to control a particular 
> > IP-packet on its way to the destination).
> > Its about a  sanity check w.r.t. the distributed view of the 
> > intra-domain network topology. Whether or not IP flows are
> > on the go is something else (they may do TTL). 
> > 
> > This sanity check may be simple and fast. And if you want, 
> > there may be different types of sanity checks:
> > some which controls just the topology without reachability 
> > info,and  without link parameters. It will be the fastest
> > (microseconds). Next there may be slower but more comprising 
> > sanity checks. At last there may even be checks on
> > proper derivatives (Routing Tables, NHLFEs,...)
> > Heinrich
> > 
> > 
> > 
> > 
> > Title: RE: Loop Detection
> > Issue is it only tells you ***somthing*** is broken. 
> > 
> > Similarly simply ensuring that the routing databases are 
> > synchronized does not perform a sanity check on the local 
> > derivative by-products (FIBs, FTN, NHLFE etc.) or the hw that 
> > implements them.
> > 
> > cheers 
> > Dave 
> > 
> > > -----Original Message----- 
> > > From: John Renwick [mailto:jkr@visi.com] 
> > > Sent: Wednesday, July 03, 2002 6:59 PM 
> > > To: Amir Hermelin; Hummel Heinrich; routing-discussion@ietf.org 
> > > Subject: RE: Loop Detection 
> > > 
> > > 
> > > Isn't TTL a very direct and immediate form of loop detection?  
> > > If you had a network tool that would illuminate the routers that 
> > > are counting a lot of packets dropped because of TTL expiration, 
> > > it could be showing you points on a routing loop. 
> > > 
> > > -- 
> > > John Renwick                          Agere Systems 
> > > Software Technical Manager            1230 Northland Drive 
> > > Phone: (651) 675-2986                 Mendota Heights, MN 55120 
> > > FAX:   (707) 897-7928                 Cell:  (612) 669-9397 
> > > mailto:jrenwick@agere.com             http://www.agere.com 
> > > 
> > > > -----Original Message----- 
> > > > From: routing-discussion-admin@ietf.org 
> > > > [mailto:routing-discussion-admin@ietf.org]On Behalf Of 
> > Amir Hermelin 
> > > > Sent: Wednesday, July 03, 2002 5:12 PM 
> > > > To: Hummel Heinrich; routing-discussion@ietf.org 
> > > > Subject: RE: Loop Detection 
> > > > 
> > > > 
> > > > my 2 cents: 
> > > > * Yes, such a tool would be very nice. Although, a 
> small portion 
> > > > of our salaries is due to the fact that routing loops are 
> > > so scary ;-) 
> > > > * The approaches would be different for types of protocols. In 
> > > > IGPs, assuming correct implementations, the protocol takes care 
> > > > of loop possibilities as long as the lspdb is in sync. When its 
> > > > not due to transient states, a tool wouldn't help since you'd 
> > > > have to run it constantly for every router's view. 
> > > > * I agree that detecting BGP mis-configuration is hard. As for 
> > > > MPLS, there are methods that are more susceptible to loops than 
> > > > others (e.g. one implementation of TE-shortcuts can 
> lead to loops 
> > > > even in a seemingly correct configuration, due to 
> metric values). 
> > > > * If this stuff moves forward, it might even mandate a 
> whole new 
> > > > workgroup. 
> > > > 
> > > > -- 
> > > > Amir Hermelin                  <mailto:amir@cwnt.com> 
> > > > Charlotte's Web Networks Inc.   <http://www.cwnt.com> 
> > > > 
> > > > 
> > > > > -----Original Message----- 
> > > > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de] 
> > > > > Sent: Tuesday, July 02, 2002 11:17 AM 
> > > > > To: routing-discussion@ietf.org 
> > > > > Subject: Re: Loop Detection 
> > > > > 
> > > > > 
> > > > > Wouldn't it be a fine goal to standardize a 
> > > tool/mechanism/protocol 
> > > > > which double-checks, within milliseconds, that all routing 
> > > > > tables of some network domain are in sync ? 
> > > > > Or at least, that the topological view, aquired by all nodes 
> > > > > of some network domain complies with the physical reality ? 
> > > > > Wouldn't this be a powerful OAM-application ? IMHO:it would 
> > > > > be better than the religious fights on OAM-MPLS. 
> > > > > 
> > > > > Heinrich Hummel 
> > > > > Siemens AG 
> > > > > 
> > > > > 
> > > > > 
> > > > >  On Tue, 4 Jun 2002, Jing Shen wrote: 
> > > > > 
> > > > > > Given a network with several routers, is there any 
> > > > > tool/application to 
> > > > > > detect loops in the routing tables? 
> > > > > 
> > > > > As far as I know, there isn't. 
> > > > > Also if the loop is due to a misconfiguration which does 
> > > not show up 
> > > > > in BGP updates, then it's harder to detect. 
> > > > > 
> > > > > -Sue 
> > 
> > _______________________________________________
> > routing-discussion mailing list
> > routing-discussion@ietf.org
> > https://www1.ietf.org/mailman/listinfo/routing-discussion
> > 
> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Thu Jul 11 04:28:25 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA12935
	for <routing-discussion-archive@odin.ietf.org>; Thu, 11 Jul 2002 04:28:25 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id EAA21124;
	Thu, 11 Jul 2002 04:22:53 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id EAA21057
	for <routing-discussion@optimus.ietf.org>; Thu, 11 Jul 2002 04:22:51 -0400 (EDT)
Received: from smtp.cwnt.com (smtp.cwnt.com [192.116.246.129])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA12717
	for <routing-discussion@ietf.org>; Thu, 11 Jul 2002 04:21:56 -0400 (EDT)
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="gb2312"
Subject: RE: Loop Detection
Date: Thu, 11 Jul 2002 11:22:19 +0300
Message-ID: <C7DF4400240AFB4095D98C7C6EC2A34A5DE16C@bart.cwnt.com>
Thread-Topic: Loop Detection
Thread-Index: AcIoewemACJ3JudZR8WKbVU/7tm3qwAN/JHw
From: "Amir Hermelin" <amir@cwnt.com>
To: <jshen@cad.zju.edu.cn>
Cc: "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id EAA21060
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

I don't think MPLS is more susceptible - it's just, being familiar with a feature such as TE shortcuts, I know for a fact some implementations' manipulation of metrics (relative, absolute) can create loop scenarios if not configured properly. And since these metrics and how they work are different than IGP, you can't really say the occurrence will be on the same level.
Generally, since MPLS has different loop detection mechanisms, than it's different than traditional routing in this aspect (as are IGP and EGP also different).

If you want, I can detail a scenario privately, but it's not the specifics that matter here.

--
Amir Hermelin                    < mailto:amir@cwnt.com>
Charlotte's Web Networks Inc.     < http://www.cwnt.com>



>-----Original Message-----
>From: Jing Shen [mailto:jshen@cad.zju.edu.cn]
>Sent: Thursday, July 11, 2002 4:34 AM
>To: Amir Hermelin
>Cc: Hummel Heinrich; routing-discussion@ietf.org
>Subject: Re: Loop Detection
>
>
>Why do you think MPLS is more susceptible to loop than other 
>technology? 
>As I think, LSP is set up by signalling along the path computed by edge
>router or
>following the items of routing tables. That is, if we just set up 
>LSPs according to routing table of IGP or EGP, the occurance of loop
>will be
>the same level of  traditional routing. 
>
>To detect loop in network, I think it's nearly impossible for a node to
>detect
>loops on some directions if it could not track routing tables at real
>time. But,
>if CR-LDP is used to set up transmission tunnel in networks, 
>the problem
>may 
>be overcomed at the stage of setup.
>
>
>
>
>Amir Hermelin wrote:
>> 
>> my 2 cents:
>> * Yes, such a tool would be very nice. Although, a small 
>portion of our salaries is due to the fact that routing loops 
>are so scary ;-)
>> * The approaches would be different for types of protocols. 
>In IGPs, assuming correct implementations, the protocol takes 
>care of loop possibilities as long as the lspdb is in sync. 
>When its not due to transient states, a tool wouldn't help 
>since you'd have to run it constantly for every router's view.
>> * I agree that detecting BGP mis-configuration is hard. As 
>for MPLS, there are methods that are more susceptible to loops 
>than others (e.g. one implementation of TE-shortcuts can lead 
>to loops even in a seemingly correct configuration, due to 
>metric values).
>> * If this stuff moves forward, it might even mandate a whole 
>new workgroup.
>> 
>> --
>> Amir Hermelin                  <mailto:amir@cwnt.com>
>> Charlotte's Web Networks Inc.   <http://www.cwnt.com>
>> 
>
>-- 
>Jing Shen
>
>Email: jshen@cad.zju.edu.cn
>
>**********************************************************************
>* The SunShine of life is made up of very little beams which is      *
>*  bright all the time                                               *
>**********************************************************************
>

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Sun Jul 14 03:40:26 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA27663
	for <routing-discussion-archive@odin.ietf.org>; Sun, 14 Jul 2002 03:40:26 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id DAA09317;
	Sun, 14 Jul 2002 03:31:15 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id DAA09292
	for <routing-discussion@optimus.ietf.org>; Sun, 14 Jul 2002 03:31:14 -0400 (EDT)
Received: from psg.com (smmsp@psg.com [147.28.0.62])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA27421
	for <routing-discussion@ietf.org>; Sun, 14 Jul 2002 03:30:18 -0400 (EDT)
Received: from localhost ([127.0.0.1] helo=127.0.0.1)
	by psg.com with esmtp (Exim 3.36 #1)
	id 17TdqZ-0004dY-00
	for routing-discussion@ietf.org; Sun, 14 Jul 2002 00:31:12 -0700
Date: Sun, 14 Jul 2002 16:30:38 +0900
From: Alex Zinin <zinin@psg.com>
X-Mailer: The Bat! (v1.51) Personal
Reply-To: Alex Zinin <zinin@psg.com>
X-Priority: 3 (Normal)
Message-ID: <13620392282.20020714163038@psg.com>
To: routing-discussion@ietf.org
Subject: RTG Area Meeting Agenda
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Folks,

 Below is the agenda for the meeting tomorrow night:

  7:30    Administriva                              5m    ADs
  7:35    AD mumbling                               15m   ADs
  7:50    WG updates                                60m   many
  8:50    BGP update                                15m   Geoff
  9:05    "Midnight Sun" routing workshop update    15m   Avri
  9:20    IPv6 address scoping                      15m   Margaret
  9:35    Open mike                                 25m
  10:00   Bar BOF

-- 
Alex



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Mon Jul 22 12:46:26 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA14584
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 12:46:26 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA27083;
	Mon, 22 Jul 2002 12:40:21 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA26972
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 12:40:14 -0400 (EDT)
Received: from gorilla.mchh.siemens.de (gorilla.mchh.siemens.de [194.138.158.18])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA13607
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 12:39:11 -0400 (EDT)
Received: from blues.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.227])
	by gorilla.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id SAA19134
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 18:40:09 +0200 (MET DST)
Received: from mchh273e.demchh201e.icn.siemens.de ([139.21.200.83])
	by blues.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id SAA01237
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 18:40:11 +0200 (MET DST)
Received: by MCHH273E with Internet Mail Service (5.5.2653.19)
	id <MY31B5RY>; Mon, 22 Jul 2002 18:40:09 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B919@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: routing-discussion@ietf.org
Subject: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 18:40:02 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

In normal, state-less IP routing we cannot do QoS-Routing, nor sneak-around the conjestion-area routing, nor traffic balanced routing. Instead, all of this is left to MPLS, where multiple states and multiple LSPs and multiple signalling is required.

However, I am pretty sure that normal state-less IP routing could do all of it much better and like to launch a discussion
on, well, Differentiated Routing. 

Additional reasons for doing it:

As far as I know, all multicast models build their delivery channels based on SPF-results. By accomplishing
QoS-routing for normal IP, QoS-routed Multicast delivery channels might consequently  come for free (or almost for free).
Also: It would prevent that further thoughts are given to Equal-Cost Multipath (ECMP) which I think is not worth the effort.

Differentiated Routing would certainly stress Differentiated Services and I think that the reason why it hasn't been 
introduced together with DiffServ is that IETF folks are only seeing the advantages of SPF routing but not its disadvantages, i.e. were therefore blocked.

2 Examples, i.e 2 analogies well-known from our roads and highways:

1) QoS Routing
Right now, there are many Europeans going on vacation by driving southbound who are willing to pay the fees for using the highways of whichever country they pass (Germany excepted) while others aren't willing. So each group has a different set of roads available to use. As a result each one will find the optimal route in accordance with what he/she likes to spend on money. 

2)
On the way south there are several hot spots which are famous for being points of conjestion. A vacationer is well advised to pull off the highway and follow the posted signs that mark asneak-around detour as to reenter further south.  
Now let's have a look at our famous rambo-zambo SPF routing paradigma. You pull off the highway.At the first junction you cry for the shortest path to your destination.Because the answer is "crankback to the highway, this option drops out, and you choose randomly any other direction. This behavior repeats at each junction. This rambo will start looping and will soon  run out of gas before he reaches the Alpes, so that the Time-To-Live is over. The smart driver however follows the detour signs, and while he is on the detouring route, he forgets about the final destination. Instead his intermediate destination is the point where he will reenter the highway. As soon as the detouring route has been done, the focus will be again on the original destination.
Isn't  this a perfect routing tunnel? the outer dest address is the re-entrance point of the autobahn,while the route to it is computed based on the fee-less set of roads.The inner destination is somewhere south at the beach and the respective forwarding may be based on computations which takes all roads and all paid highways into consideration.

So my initial question is: why can't we go after Differentiated Routing ?!
And also: why can't we go after better routing paradigmas (with respective algorithms) than SPF-routing ?!


Heinrich 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Mon Jul 22 13:39:26 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA14594
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 12:46:26 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id MAA02998
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 12:47:28 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA27083;
	Mon, 22 Jul 2002 12:40:21 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA26972
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 12:40:14 -0400 (EDT)
Received: from gorilla.mchh.siemens.de (gorilla.mchh.siemens.de [194.138.158.18])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA13607
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 12:39:11 -0400 (EDT)
Received: from blues.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.227])
	by gorilla.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id SAA19134
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 18:40:09 +0200 (MET DST)
Received: from mchh273e.demchh201e.icn.siemens.de ([139.21.200.83])
	by blues.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id SAA01237
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 18:40:11 +0200 (MET DST)
Received: by MCHH273E with Internet Mail Service (5.5.2653.19)
	id <MY31B5RY>; Mon, 22 Jul 2002 18:40:09 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B919@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: routing-discussion@ietf.org
Subject: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 18:40:02 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

In normal, state-less IP routing we cannot do QoS-Routing, nor sneak-around the conjestion-area routing, nor traffic balanced routing. Instead, all of this is left to MPLS, where multiple states and multiple LSPs and multiple signalling is required.

However, I am pretty sure that normal state-less IP routing could do all of it much better and like to launch a discussion
on, well, Differentiated Routing. 

Additional reasons for doing it:

As far as I know, all multicast models build their delivery channels based on SPF-results. By accomplishing
QoS-routing for normal IP, QoS-routed Multicast delivery channels might consequently  come for free (or almost for free).
Also: It would prevent that further thoughts are given to Equal-Cost Multipath (ECMP) which I think is not worth the effort.

Differentiated Routing would certainly stress Differentiated Services and I think that the reason why it hasn't been 
introduced together with DiffServ is that IETF folks are only seeing the advantages of SPF routing but not its disadvantages, i.e. were therefore blocked.

2 Examples, i.e 2 analogies well-known from our roads and highways:

1) QoS Routing
Right now, there are many Europeans going on vacation by driving southbound who are willing to pay the fees for using the highways of whichever country they pass (Germany excepted) while others aren't willing. So each group has a different set of roads available to use. As a result each one will find the optimal route in accordance with what he/she likes to spend on money. 

2)
On the way south there are several hot spots which are famous for being points of conjestion. A vacationer is well advised to pull off the highway and follow the posted signs that mark asneak-around detour as to reenter further south.  
Now let's have a look at our famous rambo-zambo SPF routing paradigma. You pull off the highway.At the first junction you cry for the shortest path to your destination.Because the answer is "crankback to the highway, this option drops out, and you choose randomly any other direction. This behavior repeats at each junction. This rambo will start looping and will soon  run out of gas before he reaches the Alpes, so that the Time-To-Live is over. The smart driver however follows the detour signs, and while he is on the detouring route, he forgets about the final destination. Instead his intermediate destination is the point where he will reenter the highway. As soon as the detouring route has been done, the focus will be again on the original destination.
Isn't  this a perfect routing tunnel? the outer dest address is the re-entrance point of the autobahn,while the route to it is computed based on the fee-less set of roads.The inner destination is somewhere south at the beach and the respective forwarding may be based on computations which takes all roads and all paid highways into consideration.

So my initial question is: why can't we go after Differentiated Routing ?!
And also: why can't we go after better routing paradigmas (with respective algorithms) than SPF-routing ?!


Heinrich 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 14:02:06 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA22751
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 14:02:06 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA18471;
	Mon, 22 Jul 2002 13:55:30 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA18433
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 13:55:27 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA22000
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 13:54:19 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6MHsmhJ014798;
	Mon, 22 Jul 2002 10:54:49 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020722103619.0246ba60@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Mon, 22 Jul 2002 10:54:46 -0700
To: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
From: Fred Baker <fred@cisco.com>
Subject: Re: Differentiated Routing, not only plain rambo-SPF
Cc: routing-discussion@ietf.org
In-Reply-To: <EF8E39AA846CD411BB9C00508B951F510514B919@MCHH267E>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

At 06:40 PM 7/22/2002 +0200, Hummel Heinrich wrote:
>Differentiated Routing would certainly stress Differentiated Services and 
>I think that the reason why it hasn't been introduced together with 
>DiffServ is that IETF folks are only seeing the advantages of SPF routing 
>but not its disadvantages, i.e. were therefore blocked.

The reasons are far more mundane. Routing is not only a different working 
group, it is a different area, and the work that the IETF did on the 
subject of differentiated routing in the past (mostly in OSPF) wound up not 
being important commercially and eventually got pulled out.

That said, there is some current interest in routing infrastructures that 
use both destination address and DSCP. As a result, some from the vendor 
community are looking at resurrecting and retargeting the (SPF-based) OSPF 
algorithms for that.

The key to those algorithms is not so much finding an intermediate 
destination to route to as it is to find a better metric for the traffic 
that actually results in a different and better route. TOS-based OSPF used 
the duration of a bit as the metric, and then decided whether or not to 
advertise a link as having low delay (satellite links?) or any of the other 
parameters. It would probably be better if we could actually somehow 
quantify those characteristics rather than having include/exclude lists.

>So my initial question is: why can't we go after Differentiated Routing ?!
>And also: why can't we go after better routing paradigmas (with respective 
>algorithms) than SPF-routing ?!

I should think this is largely a matter of market requirement. Demonstrate 
a market, and we can probably design something to meet its requirements. 
Absent a way to turn it into money, you might have difficulty drumming up 
commercial interest.


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Mon Jul 22 14:02:06 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA22760
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 14:02:06 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id OAA23079
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 14:03:08 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA18471;
	Mon, 22 Jul 2002 13:55:30 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA18433
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 13:55:27 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA22000
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 13:54:19 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6MHsmhJ014798;
	Mon, 22 Jul 2002 10:54:49 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020722103619.0246ba60@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Mon, 22 Jul 2002 10:54:46 -0700
To: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
From: Fred Baker <fred@cisco.com>
Subject: Re: Differentiated Routing, not only plain rambo-SPF
Cc: routing-discussion@ietf.org
In-Reply-To: <EF8E39AA846CD411BB9C00508B951F510514B919@MCHH267E>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

At 06:40 PM 7/22/2002 +0200, Hummel Heinrich wrote:
>Differentiated Routing would certainly stress Differentiated Services and 
>I think that the reason why it hasn't been introduced together with 
>DiffServ is that IETF folks are only seeing the advantages of SPF routing 
>but not its disadvantages, i.e. were therefore blocked.

The reasons are far more mundane. Routing is not only a different working 
group, it is a different area, and the work that the IETF did on the 
subject of differentiated routing in the past (mostly in OSPF) wound up not 
being important commercially and eventually got pulled out.

That said, there is some current interest in routing infrastructures that 
use both destination address and DSCP. As a result, some from the vendor 
community are looking at resurrecting and retargeting the (SPF-based) OSPF 
algorithms for that.

The key to those algorithms is not so much finding an intermediate 
destination to route to as it is to find a better metric for the traffic 
that actually results in a different and better route. TOS-based OSPF used 
the duration of a bit as the metric, and then decided whether or not to 
advertise a link as having low delay (satellite links?) or any of the other 
parameters. It would probably be better if we could actually somehow 
quantify those characteristics rather than having include/exclude lists.

>So my initial question is: why can't we go after Differentiated Routing ?!
>And also: why can't we go after better routing paradigmas (with respective 
>algorithms) than SPF-routing ?!

I should think this is largely a matter of market requirement. Demonstrate 
a market, and we can probably design something to meet its requirements. 
Absent a way to turn it into money, you might have difficulty drumming up 
commercial interest.


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 14:04:36 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA23046
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 14:04:36 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA19558;
	Mon, 22 Jul 2002 13:57:31 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA19506
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 13:57:27 -0400 (EDT)
Received: from mailgate.pit.comms.marconi.com (mailgate.pit.comms.marconi.com [169.144.68.6])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA22214
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 13:56:24 -0400 (EDT)
Received: from mailman.pit.comms.marconi.com (mailman.pit.comms.marconi.com [169.144.2.12])
	by mailgate.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id NAA21317;
	Mon, 22 Jul 2002 13:56:53 -0400 (EDT)
Received: from whq-msgrtr-01.pit.comms.marconi.com (whq-msgrtr-01.pit.comms.marconi.com [169.144.2.221])
	by mailman.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id NAA04873;
	Mon, 22 Jul 2002 13:56:55 -0400 (EDT)
Received: by whq-msgrtr-01.pit.comms.marconi.com with Internet Mail Service (5.5.2650.21)
	id <3W9883S7>; Mon, 22 Jul 2002 13:56:54 -0400
Message-ID: <39469E08BD83D411A3D900204840EC55763270@vie-msgusr-01.dc.fore.com>
From: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
To: "'Hummel Heinrich'" <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 13:56:52 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hummel,

  Very interesting topic. Let me give my view.

  First of all, let me put what you described in a different way.

  Different types of services (call them ToS, CoS, GoS what ever)
  exists and support for those are well defined in the IP history
  since its existence (thanks for those who envisioned & reserved
  few bits in IP header). Those are TOS bits, and almost all
  of the IP routing protocols (at least OSPF) supports TOS.

  In other words, OSPF has different metrics for different
  TOS (min delay, max throughput, max reliability, min cost etc)
  You can (if you would like to) do SPF based on specific TOS
  metrics advertised into OSPF. That solves your first problem
  based on roads & highways.

  TOS and TOS-Routing has been deprecated in the Internet
  because of some historical & deployment reasons. 
  TOS has been changed to support DiffServ. But unfortunately
  TOS-Routing is not changed to Diff-Routing. I really
  don't know why? Pioneers can answer this question.
  But in my opinion, Yes! Diff-Routing solves some of the
  know problems you described.

  MPLS-TE on the other hand is based on Source Routing (atleast
  for now). The envisioned extensions will eventually lead to 
  Diff-Routing (and even better). For example, multiple metrics 
  for DiffServ have been proposed to IETF.
  (see draft-fedyk-isis-ospf-te-metrics-01.txt)
  
  I thought of proposing use of OSPF TOS metrics for DiffServ
  Routing (instead of above extensions) but I don't know
  how war this will be useful...

--
Venkata.

-> -----Original Message-----
-> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
-> Sent: Monday, July 22, 2002 12:40 PM
-> To: routing-discussion@ietf.org
-> Subject: Differentiated Routing, not only plain rambo-SPF
-> 
-> 
-> In normal, state-less IP routing we cannot do QoS-Routing, 
-> nor sneak-around the conjestion-area routing, nor traffic 
-> balanced routing. Instead, all of this is left to MPLS, 
-> where multiple states and multiple LSPs and multiple 
-> signalling is required.
-> 
-> However, I am pretty sure that normal state-less IP routing 
-> could do all of it much better and like to launch a discussion
-> on, well, Differentiated Routing. 
-> 
-> Additional reasons for doing it:
-> 
-> As far as I know, all multicast models build their delivery 
-> channels based on SPF-results. By accomplishing
-> QoS-routing for normal IP, QoS-routed Multicast delivery 
-> channels might consequently  come for free (or almost for free).
-> Also: It would prevent that further thoughts are given to 
-> Equal-Cost Multipath (ECMP) which I think is not worth the effort.
-> 
-> Differentiated Routing would certainly stress Differentiated 
-> Services and I think that the reason why it hasn't been 
-> introduced together with DiffServ is that IETF folks are 
-> only seeing the advantages of SPF routing but not its 
-> disadvantages, i.e. were therefore blocked.
-> 
-> 2 Examples, i.e 2 analogies well-known from our roads and highways:
-> 
-> 1) QoS Routing
-> Right now, there are many Europeans going on vacation by 
-> driving southbound who are willing to pay the fees for using 
-> the highways of whichever country they pass (Germany 
-> excepted) while others aren't willing. So each group has a 
-> different set of roads available to use. As a result each 
-> one will find the optimal route in accordance with what 
-> he/she likes to spend on money. 
-> 
-> 2)
-> On the way south there are several hot spots which are 
-> famous for being points of conjestion. A vacationer is well 
-> advised to pull off the highway and follow the posted signs 
-> that mark asneak-around detour as to reenter further south.  
-> Now let's have a look at our famous rambo-zambo SPF routing 
-> paradigma. You pull off the highway.At the first junction 
-> you cry for the shortest path to your destination.Because 
-> the answer is "crankback to the highway, this option drops 
-> out, and you choose randomly any other direction. This 
-> behavior repeats at each junction. This rambo will start 
-> looping and will soon  run out of gas before he reaches the 
-> Alpes, so that the Time-To-Live is over. The smart driver 
-> however follows the detour signs, and while he is on the 
-> detouring route, he forgets about the final destination. 
-> Instead his intermediate destination is the point where he 
-> will reenter the highway. As soon as the detouring route has 
-> been done, the focus will be again on the original destination.
-> Isn't  this a perfect routing tunnel? the outer dest address 
-> is the re-entrance point of the autobahn,while the route to 
-> it is computed based on the fee-less set of roads.The inner 
-> destination is somewhere south at the beach and the 
-> respective forwarding may be based on computations which 
-> takes all roads and all paid highways into consideration.
-> 
-> So my initial question is: why can't we go after 
-> Differentiated Routing ?!
-> And also: why can't we go after better routing paradigmas 
-> (with respective algorithms) than SPF-routing ?!
-> 
-> 
-> Heinrich 
-> 
-> _______________________________________________
-> routing-discussion mailing list
-> routing-discussion@ietf.org
-> https://www1.ietf.org/mailman/listinfo/routing-discussion
-> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Mon Jul 22 14:04:37 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA23059
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 14:04:37 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id OAA24458
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 14:05:39 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA19558;
	Mon, 22 Jul 2002 13:57:31 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA19506
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 13:57:27 -0400 (EDT)
Received: from mailgate.pit.comms.marconi.com (mailgate.pit.comms.marconi.com [169.144.68.6])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA22214
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 13:56:24 -0400 (EDT)
Received: from mailman.pit.comms.marconi.com (mailman.pit.comms.marconi.com [169.144.2.12])
	by mailgate.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id NAA21317;
	Mon, 22 Jul 2002 13:56:53 -0400 (EDT)
Received: from whq-msgrtr-01.pit.comms.marconi.com (whq-msgrtr-01.pit.comms.marconi.com [169.144.2.221])
	by mailman.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id NAA04873;
	Mon, 22 Jul 2002 13:56:55 -0400 (EDT)
Received: by whq-msgrtr-01.pit.comms.marconi.com with Internet Mail Service (5.5.2650.21)
	id <3W9883S7>; Mon, 22 Jul 2002 13:56:54 -0400
Message-ID: <39469E08BD83D411A3D900204840EC55763270@vie-msgusr-01.dc.fore.com>
From: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
To: "'Hummel Heinrich'" <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 13:56:52 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hummel,

  Very interesting topic. Let me give my view.

  First of all, let me put what you described in a different way.

  Different types of services (call them ToS, CoS, GoS what ever)
  exists and support for those are well defined in the IP history
  since its existence (thanks for those who envisioned & reserved
  few bits in IP header). Those are TOS bits, and almost all
  of the IP routing protocols (at least OSPF) supports TOS.

  In other words, OSPF has different metrics for different
  TOS (min delay, max throughput, max reliability, min cost etc)
  You can (if you would like to) do SPF based on specific TOS
  metrics advertised into OSPF. That solves your first problem
  based on roads & highways.

  TOS and TOS-Routing has been deprecated in the Internet
  because of some historical & deployment reasons. 
  TOS has been changed to support DiffServ. But unfortunately
  TOS-Routing is not changed to Diff-Routing. I really
  don't know why? Pioneers can answer this question.
  But in my opinion, Yes! Diff-Routing solves some of the
  know problems you described.

  MPLS-TE on the other hand is based on Source Routing (atleast
  for now). The envisioned extensions will eventually lead to 
  Diff-Routing (and even better). For example, multiple metrics 
  for DiffServ have been proposed to IETF.
  (see draft-fedyk-isis-ospf-te-metrics-01.txt)
  
  I thought of proposing use of OSPF TOS metrics for DiffServ
  Routing (instead of above extensions) but I don't know
  how war this will be useful...

--
Venkata.

-> -----Original Message-----
-> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
-> Sent: Monday, July 22, 2002 12:40 PM
-> To: routing-discussion@ietf.org
-> Subject: Differentiated Routing, not only plain rambo-SPF
-> 
-> 
-> In normal, state-less IP routing we cannot do QoS-Routing, 
-> nor sneak-around the conjestion-area routing, nor traffic 
-> balanced routing. Instead, all of this is left to MPLS, 
-> where multiple states and multiple LSPs and multiple 
-> signalling is required.
-> 
-> However, I am pretty sure that normal state-less IP routing 
-> could do all of it much better and like to launch a discussion
-> on, well, Differentiated Routing. 
-> 
-> Additional reasons for doing it:
-> 
-> As far as I know, all multicast models build their delivery 
-> channels based on SPF-results. By accomplishing
-> QoS-routing for normal IP, QoS-routed Multicast delivery 
-> channels might consequently  come for free (or almost for free).
-> Also: It would prevent that further thoughts are given to 
-> Equal-Cost Multipath (ECMP) which I think is not worth the effort.
-> 
-> Differentiated Routing would certainly stress Differentiated 
-> Services and I think that the reason why it hasn't been 
-> introduced together with DiffServ is that IETF folks are 
-> only seeing the advantages of SPF routing but not its 
-> disadvantages, i.e. were therefore blocked.
-> 
-> 2 Examples, i.e 2 analogies well-known from our roads and highways:
-> 
-> 1) QoS Routing
-> Right now, there are many Europeans going on vacation by 
-> driving southbound who are willing to pay the fees for using 
-> the highways of whichever country they pass (Germany 
-> excepted) while others aren't willing. So each group has a 
-> different set of roads available to use. As a result each 
-> one will find the optimal route in accordance with what 
-> he/she likes to spend on money. 
-> 
-> 2)
-> On the way south there are several hot spots which are 
-> famous for being points of conjestion. A vacationer is well 
-> advised to pull off the highway and follow the posted signs 
-> that mark asneak-around detour as to reenter further south.  
-> Now let's have a look at our famous rambo-zambo SPF routing 
-> paradigma. You pull off the highway.At the first junction 
-> you cry for the shortest path to your destination.Because 
-> the answer is "crankback to the highway, this option drops 
-> out, and you choose randomly any other direction. This 
-> behavior repeats at each junction. This rambo will start 
-> looping and will soon  run out of gas before he reaches the 
-> Alpes, so that the Time-To-Live is over. The smart driver 
-> however follows the detour signs, and while he is on the 
-> detouring route, he forgets about the final destination. 
-> Instead his intermediate destination is the point where he 
-> will reenter the highway. As soon as the detouring route has 
-> been done, the focus will be again on the original destination.
-> Isn't  this a perfect routing tunnel? the outer dest address 
-> is the re-entrance point of the autobahn,while the route to 
-> it is computed based on the fee-less set of roads.The inner 
-> destination is somewhere south at the beach and the 
-> respective forwarding may be based on computations which 
-> takes all roads and all paid highways into consideration.
-> 
-> So my initial question is: why can't we go after 
-> Differentiated Routing ?!
-> And also: why can't we go after better routing paradigmas 
-> (with respective algorithms) than SPF-routing ?!
-> 
-> 
-> Heinrich 
-> 
-> _______________________________________________
-> routing-discussion mailing list
-> routing-discussion@ietf.org
-> https://www1.ietf.org/mailman/listinfo/routing-discussion
-> 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 14:52:35 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27559
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 14:52:35 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA15008;
	Mon, 22 Jul 2002 14:43:51 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA14953
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 14:43:47 -0400 (EDT)
Received: from mailgate.pit.comms.marconi.com (mailgate.pit.comms.marconi.com [169.144.68.6])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA26558
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 14:42:46 -0400 (EDT)
Received: from mailman.pit.comms.marconi.com (mailman.pit.comms.marconi.com [169.144.2.12])
	by mailgate.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA25076;
	Mon, 22 Jul 2002 14:43:14 -0400 (EDT)
Received: from whq-msgrtr-01.pit.comms.marconi.com (whq-msgrtr-01.pit.comms.marconi.com [169.144.2.221])
	by mailman.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA17302;
	Mon, 22 Jul 2002 14:43:14 -0400 (EDT)
Received: by whq-msgrtr-01.pit.comms.marconi.com with Internet Mail Service (5.5.2650.21)
	id <3W988SHM>; Mon, 22 Jul 2002 14:43:13 -0400
Message-ID: <39469E08BD83D411A3D900204840EC55763272@vie-msgusr-01.dc.fore.com>
From: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
To: "'Fred Baker'" <fred@cisco.com>,
        Hummel Heinrich
	 <Heinrich.Hummel@icn.siemens.de>
Cc: routing-discussion@ietf.org
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 14:43:04 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Fred,

  I have more questions on the future Internet QoS
  Routing (answer/correct me if any of my views are
  impractical).

  As I answered before, my worry is not DiffServ 
  Routing. The on-going extensions will eventually
  lead to better routing.

  But that is very limited to an Area (a small fraction
  in the Internet). Using all these QoS routing extensions
  confined to these small areas will not buy us (an end
  users) anything in the long run. I am speaking
  in the interest of end-to-end Internet world.

  We have an IGP/EGP split and the interoperation
  of these to carry QoS requirements are not at all 
  triggered. The current signaling protocols are also
  very stringent and won't extend end-to-end (not even
  an AS). With out these:

  "How can we deliver a future real-time services 
   (given a SIP call) end-to-end with some guarantee
   over the Internet?"

  "Do you think that there is no need for such 
   Hierarchical QoS Routing?"
  
  I know that, NSIS is trying to solve signaling part 
  of the problem. But signaling follows what routing
  dictates. How are we going to solve the problem?

  If we worry about scalability & aggregation of
  hierarchical QoS routing, we can search for other
  possibilities. For example, "LSP stitching" from
  source AS till destination address AS. This does
  not require any enhancements to IGPs but requires
  some way of communications of the QoS parameters
  across EGP boundaries. Do you agree?

  PS: I am speaking as an individual. No prejudices
      towards any union/task/forum. I just would like
      to evaluate evolution :)
  
--
Venkata.

-> >So my initial question is: why can't we go after 
-> Differentiated Routing ?!
-> >And also: why can't we go after better routing paradigmas 
-> (with respective 
-> >algorithms) than SPF-routing ?!
-> 
-> I should think this is largely a matter of market 
-> requirement. Demonstrate 
-> a market, and we can probably design something to meet its 
-> requirements. 
-> Absent a way to turn it into money, you might have 
-> difficulty drumming up 
-> commercial interest.

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Mon Jul 22 14:52:36 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27573
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 14:52:36 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id OAA19976
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 14:53:37 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA15008;
	Mon, 22 Jul 2002 14:43:51 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA14953
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 14:43:47 -0400 (EDT)
Received: from mailgate.pit.comms.marconi.com (mailgate.pit.comms.marconi.com [169.144.68.6])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA26558
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 14:42:46 -0400 (EDT)
Received: from mailman.pit.comms.marconi.com (mailman.pit.comms.marconi.com [169.144.2.12])
	by mailgate.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA25076;
	Mon, 22 Jul 2002 14:43:14 -0400 (EDT)
Received: from whq-msgrtr-01.pit.comms.marconi.com (whq-msgrtr-01.pit.comms.marconi.com [169.144.2.221])
	by mailman.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA17302;
	Mon, 22 Jul 2002 14:43:14 -0400 (EDT)
Received: by whq-msgrtr-01.pit.comms.marconi.com with Internet Mail Service (5.5.2650.21)
	id <3W988SHM>; Mon, 22 Jul 2002 14:43:13 -0400
Message-ID: <39469E08BD83D411A3D900204840EC55763272@vie-msgusr-01.dc.fore.com>
From: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
To: "'Fred Baker'" <fred@cisco.com>,
        Hummel Heinrich
	 <Heinrich.Hummel@icn.siemens.de>
Cc: routing-discussion@ietf.org
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 14:43:04 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Fred,

  I have more questions on the future Internet QoS
  Routing (answer/correct me if any of my views are
  impractical).

  As I answered before, my worry is not DiffServ 
  Routing. The on-going extensions will eventually
  lead to better routing.

  But that is very limited to an Area (a small fraction
  in the Internet). Using all these QoS routing extensions
  confined to these small areas will not buy us (an end
  users) anything in the long run. I am speaking
  in the interest of end-to-end Internet world.

  We have an IGP/EGP split and the interoperation
  of these to carry QoS requirements are not at all 
  triggered. The current signaling protocols are also
  very stringent and won't extend end-to-end (not even
  an AS). With out these:

  "How can we deliver a future real-time services 
   (given a SIP call) end-to-end with some guarantee
   over the Internet?"

  "Do you think that there is no need for such 
   Hierarchical QoS Routing?"
  
  I know that, NSIS is trying to solve signaling part 
  of the problem. But signaling follows what routing
  dictates. How are we going to solve the problem?

  If we worry about scalability & aggregation of
  hierarchical QoS routing, we can search for other
  possibilities. For example, "LSP stitching" from
  source AS till destination address AS. This does
  not require any enhancements to IGPs but requires
  some way of communications of the QoS parameters
  across EGP boundaries. Do you agree?

  PS: I am speaking as an individual. No prejudices
      towards any union/task/forum. I just would like
      to evaluate evolution :)
  
--
Venkata.

-> >So my initial question is: why can't we go after 
-> Differentiated Routing ?!
-> >And also: why can't we go after better routing paradigmas 
-> (with respective 
-> >algorithms) than SPF-routing ?!
-> 
-> I should think this is largely a matter of market 
-> requirement. Demonstrate 
-> a market, and we can probably design something to meet its 
-> requirements. 
-> Absent a way to turn it into money, you might have 
-> difficulty drumming up 
-> commercial interest.

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 15:43:20 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA02361
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 15:43:20 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA08946;
	Mon, 22 Jul 2002 15:32:23 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA08890
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 15:32:19 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA01114
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 15:31:17 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6MJVhhJ029177;
	Mon, 22 Jul 2002 12:31:43 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020722120742.0248f1e8@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Mon, 22 Jul 2002 12:31:40 -0700
To: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
From: Fred Baker <fred@cisco.com>
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Cc: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
In-Reply-To: <39469E08BD83D411A3D900204840EC55763272@vie-msgusr-01.dc.fo
 re.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org


>PS: I am speaking as an individual.

This is an IETF list. We are all individuals in this forum. Frankly, if 
your opinion would be different if you posted from a different email 
address, it's not actually your opinion, and I'm not particularly 
interested in it.

>I know that, NSIS is trying to solve signaling part of the problem. But 
>signaling follows what routing dictates. How are we going to solve the problem?

I guess my question is "please state the problem". You and Hummel would 
like for the solution to include scalable QoS routing, for some definition 
of that term, and Hummel's not happy with distributed SPF as an algorithm. 
He comments that it should be related to routing at the network layer; your 
answer suggests that it should be routing at a circuit layer, and should 
include some concept of routing (SIP) calls. What is most interesting, at 
least to me, is "please spell out the problem that this solution is 
intended to solve?"

I mentioned destination+DSCP routing. The issues that is intended to be a 
solution to come from two general categories of user requests -

         "I would like to route my voice traffic differently than my data 
traffic in a
          converged network"
and
         "My network is composed of a large number of very low bandwidth 
high delay
          facilities. I would like to use them all simultaneously to best 
advantage
          if at all possible."

Note that neither is a complaint about delays or bandwidth per se. The 
former is more a policy question - "I have specific links for that traffic 
and I want to use them". The latter might be restated "do what I mean", so 
any approach to it is likely to partially succeed and partially fail. What 
the target audience really needs to do is get more bandwidth, but it's not 
something readily done.

I'm also interested in routing that uses measured delay as a metric, in 
OSPF or BGP. The problem statement as put by a variety of companies such as 
Sockeye or NetVMG might be stated "in a multihomed situation, select the 
service provider that will give you the best round trip time on a request." 
The problem statement as put by Akamai and friends might be "select a route 
or server that will give you the highest delivery rate using TCP." But 
again, I don't see anything related to that in either your extended LSPs or 
Hummel's issues with SPF, though.

What is the problem you want to solve, that your proposal might be a 
solution to?


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@ns.ietf.org  Mon Jul 22 15:43:21 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA02378
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 15:43:21 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id PAA15426
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 15:44:22 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA08946;
	Mon, 22 Jul 2002 15:32:23 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA08890
	for <routing-discussion@optimus.ietf.org>; Mon, 22 Jul 2002 15:32:19 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA01114
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 15:31:17 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6MJVhhJ029177;
	Mon, 22 Jul 2002 12:31:43 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020722120742.0248f1e8@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Mon, 22 Jul 2002 12:31:40 -0700
To: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
From: Fred Baker <fred@cisco.com>
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Cc: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>,
        routing-discussion@ietf.org
In-Reply-To: <39469E08BD83D411A3D900204840EC55763272@vie-msgusr-01.dc.fo
 re.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org


>PS: I am speaking as an individual.

This is an IETF list. We are all individuals in this forum. Frankly, if 
your opinion would be different if you posted from a different email 
address, it's not actually your opinion, and I'm not particularly 
interested in it.

>I know that, NSIS is trying to solve signaling part of the problem. But 
>signaling follows what routing dictates. How are we going to solve the problem?

I guess my question is "please state the problem". You and Hummel would 
like for the solution to include scalable QoS routing, for some definition 
of that term, and Hummel's not happy with distributed SPF as an algorithm. 
He comments that it should be related to routing at the network layer; your 
answer suggests that it should be routing at a circuit layer, and should 
include some concept of routing (SIP) calls. What is most interesting, at 
least to me, is "please spell out the problem that this solution is 
intended to solve?"

I mentioned destination+DSCP routing. The issues that is intended to be a 
solution to come from two general categories of user requests -

         "I would like to route my voice traffic differently than my data 
traffic in a
          converged network"
and
         "My network is composed of a large number of very low bandwidth 
high delay
          facilities. I would like to use them all simultaneously to best 
advantage
          if at all possible."

Note that neither is a complaint about delays or bandwidth per se. The 
former is more a policy question - "I have specific links for that traffic 
and I want to use them". The latter might be restated "do what I mean", so 
any approach to it is likely to partially succeed and partially fail. What 
the target audience really needs to do is get more bandwidth, but it's not 
something readily done.

I'm also interested in routing that uses measured delay as a metric, in 
OSPF or BGP. The problem statement as put by a variety of companies such as 
Sockeye or NetVMG might be stated "in a multihomed situation, select the 
service provider that will give you the best round trip time on a request." 
The problem statement as put by Akamai and friends might be "select a route 
or server that will give you the highest delivery rate using TCP." But 
again, I don't see anything related to that in either your extended LSPs or 
Hummel's issues with SPF, though.

What is the problem you want to solve, that your proposal might be a 
solution to?


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 17:24:57 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA10930
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:24:57 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA27408;
	Mon, 22 Jul 2002 17:10:49 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA27361
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 17:10:46 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA09571
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 17:09:45 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6MLABhJ007564;
	Mon, 22 Jul 2002 14:10:11 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020722134321.02414278@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Mon, 22 Jul 2002 14:10:10 -0700
To: svenkata@mail.gmu.edu
From: Fred Baker <fred@cisco.com>
Subject: Re: End-to-End QoS Routing
Cc: routing-discussion@ietf.org
In-Reply-To: <b96485b9316a.b9316ab96485@gmu.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

At 04:24 PM 7/22/2002 -0400, svenkata@gmu.edu wrote:
>   You completely misunderstood my good intensions.

Sorry you felt that way. I wasn't questioning your intentions.

>   Question:
>   All the current QoS routing (MPLS-TE)

I'm not convinced that all QoS routing is in MPLS-TE. That is simply what 
is visible as IETF effort.

>Using all these QoS routing extensions are confined to these small areas 
>will not buy us (an end users) anything in the long run (end-to-end 
>Internet perspective). Do you agree?

That's a *really* open-ended question.

Traffic engineering is not really there to help you, the end user; it is 
there to help the service provider give a service to its customers (not you 
personally, the company that contracts service on your behalf) in a 
consistent and cost-effective manner. If you are having a problem with 
inconsistent service, having your service provider do some traffic 
engineering might help, and might not; if you're not having that problem, 
all the traffic engineering in the world won't help with whatever problems 
you do have.

>   If not,
>   "How can we deliver real-time services (differentiated)
>    with some end-to-end guarantee over the Internet?"

Again, exactly what *problem* do you need to solve? Do you need 
differentiation? What guarantees do you need that you are not getting now?

>   Possible Solution(s):

Until I have a problem statement, it's really hard to comment on potential 
solutions. I gave you several problem statements that I am dealing with. 
Give me yours?


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@ns.ietf.org  Mon Jul 22 17:24:57 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA10938
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:24:57 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id RAA05366
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 17:25:57 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA27408;
	Mon, 22 Jul 2002 17:10:49 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA27361
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 17:10:46 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA09571
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 17:09:45 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6MLABhJ007564;
	Mon, 22 Jul 2002 14:10:11 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020722134321.02414278@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Mon, 22 Jul 2002 14:10:10 -0700
To: svenkata@mail.gmu.edu
From: Fred Baker <fred@cisco.com>
Subject: Re: End-to-End QoS Routing
Cc: routing-discussion@ietf.org
In-Reply-To: <b96485b9316a.b9316ab96485@gmu.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

At 04:24 PM 7/22/2002 -0400, svenkata@gmu.edu wrote:
>   You completely misunderstood my good intensions.

Sorry you felt that way. I wasn't questioning your intentions.

>   Question:
>   All the current QoS routing (MPLS-TE)

I'm not convinced that all QoS routing is in MPLS-TE. That is simply what 
is visible as IETF effort.

>Using all these QoS routing extensions are confined to these small areas 
>will not buy us (an end users) anything in the long run (end-to-end 
>Internet perspective). Do you agree?

That's a *really* open-ended question.

Traffic engineering is not really there to help you, the end user; it is 
there to help the service provider give a service to its customers (not you 
personally, the company that contracts service on your behalf) in a 
consistent and cost-effective manner. If you are having a problem with 
inconsistent service, having your service provider do some traffic 
engineering might help, and might not; if you're not having that problem, 
all the traffic engineering in the world won't help with whatever problems 
you do have.

>   If not,
>   "How can we deliver real-time services (differentiated)
>    with some end-to-end guarantee over the Internet?"

Again, exactly what *problem* do you need to solve? Do you need 
differentiation? What guarantees do you need that you are not getting now?

>   Possible Solution(s):

Until I have a problem statement, it's really hard to comment on potential 
solutions. I gave you several problem statements that I am dealing with. 
Give me yours?


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 17:31:35 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA11506
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:31:35 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01107;
	Mon, 22 Jul 2002 17:18:05 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01063
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 17:18:02 -0400 (EDT)
Received: from kc-msxproto2.kc.umkc.edu (kc-msxproto2.kc.umkc.edu [134.193.143.159])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA10138
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 17:17:01 -0400 (EDT)
Received: from KC-MAIL2.kc.umkc.edu ([134.193.143.162] RDNS failed) by kc-msxproto2.kc.umkc.edu with Microsoft SMTPSVC(5.0.2195.4905);
	 Mon, 22 Jul 2002 16:17:59 -0500
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 16:17:20 -0500
Message-ID: <A29143A40AEAE3459FF829D1DD4331613C24F2@KC-MAIL2.kc.umkc.edu>
Thread-Topic: Differentiated Routing, not only plain rambo-SPF
Thread-Index: AcIxomqvZORnaMu1SlW/dN1vQw4RiQAHeeQD
From: "Ayyasamy, Senthilkumar  (UMKC-Student)" <saq66@umkc.edu>
To: "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
X-OriginalArrivalTime: 22 Jul 2002 21:17:59.0754 (UTC) FILETIME=[4187EAA0:01C231C5]
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from base64 to 8bit by optimus.ietf.org id RAA01064
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

>Differentiated Routing would certainly stress Differentiated Services 
>and I think that the reason why it hasn't been
>introduced together with DiffServ is that IETF folks are only 
>seeing the advantages of SPF routing but not its 
>disadvantages, i.e. were therefore blocked.
I have studied something reverse in some cases
(with particular reference to premium service in diffserv)
An excerpt from a recent infcomm paper,
* Originally, the DiffServ scheme is decoupled from IP routing 
intentionally, meaning that all traffic 
between each source-destination pair follows the same path no 
matter which service class it belongs to and DiffServ itself has no
effect on IP routing decisions. Due to premium trafficâs high priority
this could lead to some problems for the low-priority traffic when the
 volume of premium class traffic is high. That is to say,without taking 
routing into consideration, the premium class traffic imposes very 
negative influences on other classes of traffic, especially when the 
network is highly loaded.*
The authors modify SPF algo for avoiding problems in diffserv networks.
http://www.ieee-infocom.org/2002/papers/675.pdf
Though the paper tries to find solution for something which have been
done already (inter-class effects are discussed in the orginal EF 
RFC itself,) the authors have come up with good algebra and routing
problem analysis in diffserv networks which will help further research
in coupling IP routing and diffserv.
 
>So my initial question is: why can't we go after Differentiated Routing ?!
>And also: why can't we go after better routing paradigmas (with respective >algorithms) than SPF-routing ?!
 
As said, recent interest is using OSPF metric optimization for network planning.
People tweak around OSPF weights for managing their network. Hence there 
should be a strong reason for moving from the present routing paradigm. I am 
wary of many stateful routing schemes deployment.
 
 
 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@ns.ietf.org  Mon Jul 22 17:31:35 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA11515
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:31:35 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id RAA08897
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 17:32:35 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01107;
	Mon, 22 Jul 2002 17:18:05 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01063
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 17:18:02 -0400 (EDT)
Received: from kc-msxproto2.kc.umkc.edu (kc-msxproto2.kc.umkc.edu [134.193.143.159])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA10138
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 17:17:01 -0400 (EDT)
Received: from KC-MAIL2.kc.umkc.edu ([134.193.143.162] RDNS failed) by kc-msxproto2.kc.umkc.edu with Microsoft SMTPSVC(5.0.2195.4905);
	 Mon, 22 Jul 2002 16:17:59 -0500
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 16:17:20 -0500
Message-ID: <A29143A40AEAE3459FF829D1DD4331613C24F2@KC-MAIL2.kc.umkc.edu>
Thread-Topic: Differentiated Routing, not only plain rambo-SPF
Thread-Index: AcIxomqvZORnaMu1SlW/dN1vQw4RiQAHeeQD
From: "Ayyasamy, Senthilkumar  (UMKC-Student)" <saq66@umkc.edu>
To: "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
X-OriginalArrivalTime: 22 Jul 2002 21:17:59.0754 (UTC) FILETIME=[4187EAA0:01C231C5]
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from base64 to 8bit by optimus.ietf.org id RAA01064
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

>Differentiated Routing would certainly stress Differentiated Services 
>and I think that the reason why it hasn't been
>introduced together with DiffServ is that IETF folks are only 
>seeing the advantages of SPF routing but not its 
>disadvantages, i.e. were therefore blocked.
I have studied something reverse in some cases
(with particular reference to premium service in diffserv)
An excerpt from a recent infcomm paper,
* Originally, the DiffServ scheme is decoupled from IP routing 
intentionally, meaning that all traffic 
between each source-destination pair follows the same path no 
matter which service class it belongs to and DiffServ itself has no
effect on IP routing decisions. Due to premium trafficâs high priority
this could lead to some problems for the low-priority traffic when the
 volume of premium class traffic is high. That is to say,without taking 
routing into consideration, the premium class traffic imposes very 
negative influences on other classes of traffic, especially when the 
network is highly loaded.*
The authors modify SPF algo for avoiding problems in diffserv networks.
http://www.ieee-infocom.org/2002/papers/675.pdf
Though the paper tries to find solution for something which have been
done already (inter-class effects are discussed in the orginal EF 
RFC itself,) the authors have come up with good algebra and routing
problem analysis in diffserv networks which will help further research
in coupling IP routing and diffserv.
 
>So my initial question is: why can't we go after Differentiated Routing ?!
>And also: why can't we go after better routing paradigmas (with respective >algorithms) than SPF-routing ?!
 
As said, recent interest is using OSPF metric optimization for network planning.
People tweak around OSPF weights for managing their network. Hence there 
should be a strong reason for moving from the present routing paradigm. I am 
wary of many stateful routing schemes deployment.
 
 
 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 17:32:43 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA11600
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:32:42 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01681;
	Mon, 22 Jul 2002 17:19:09 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01619
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 17:19:05 -0400 (EDT)
Received: from kc-msxproto2.kc.umkc.edu (kc-msxproto2.kc.umkc.edu [134.193.143.159])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA10229
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 17:18:04 -0400 (EDT)
Received: from KC-MAIL2.kc.umkc.edu ([134.193.143.162] RDNS failed) by kc-msxproto2.kc.umkc.edu with Microsoft SMTPSVC(5.0.2195.4905);
	 Mon, 22 Jul 2002 16:19:05 -0500
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 16:19:04 -0500
Message-ID: <A29143A40AEAE3459FF829D1DD4331613C24F3@KC-MAIL2.kc.umkc.edu>
Thread-Topic: Differentiated Routing, not only plain rambo-SPF
Thread-Index: AcIxomqvZORnaMu1SlW/dN1vQw4RiQAItVPP
From: "Ayyasamy, Senthilkumar  (UMKC-Student)" <saq66@umkc.edu>
To: "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
X-OriginalArrivalTime: 22 Jul 2002 21:19:05.0271 (UTC) FILETIME=[68950470:01C231C5]
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from base64 to 8bit by optimus.ietf.org id RAA01620
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

>Differentiated Routing would certainly stress Differentiated Services 
>and I think that the reason why it hasn't been
>introduced together with DiffServ is that IETF folks are only 
>seeing the advantages of SPF routing but not its 
>disadvantages, i.e. were therefore blocked.
I have studied something reverse in some cases
(with particular reference to premium service in diffserv)
 
An excerpt from a recent infcomm paper,
* Originally, the DiffServ scheme is decoupled from IP routing 
intentionally, meaning that all traffic 
between each source-destination pair follows the same path no 
matter which service class it belongs to and DiffServ itself has no
effect on IP routing decisions. Due to premium trafficâs high priority
this could lead to some problems for the low-priority traffic when the
 volume of premium class traffic is high. That is to say,without taking 
routing into consideration, the premium class traffic imposes very 
negative influences on other classes of traffic, especially when the 
network is highly loaded.*
The authors modify SPF algo for avoiding problems in diffserv networks.
 
http://www.ieee-infocom.org/2002/papers/675.pdf
Though the paper tries to find solution for something which have been
done already (inter-class effects are discussed in the orginal EF 
RFC itself,) the authors have come up with good algebra and routing
problem analysis in diffserv networks which will help further research
in coupling IP routing and diffserv.
 
>So my initial question is: why can't we go after Differentiated Routing ?!
>And also: why can't we go after better routing paradigmas (with respective >algorithms) than SPF-routing ?!
 
As said, recent interest is using OSPF metric optimization for network planning.
People tweak around OSPF weights for managing their network. Hence there 
should be a strong reason for moving from the present routing paradigm. I am 
wary of many stateful routing schemes deployment.
 
 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@ns.ietf.org  Mon Jul 22 17:32:44 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA11613
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:32:44 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id RAA09499
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 17:33:43 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01681;
	Mon, 22 Jul 2002 17:19:09 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA01619
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 17:19:05 -0400 (EDT)
Received: from kc-msxproto2.kc.umkc.edu (kc-msxproto2.kc.umkc.edu [134.193.143.159])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA10229
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 17:18:04 -0400 (EDT)
Received: from KC-MAIL2.kc.umkc.edu ([134.193.143.162] RDNS failed) by kc-msxproto2.kc.umkc.edu with Microsoft SMTPSVC(5.0.2195.4905);
	 Mon, 22 Jul 2002 16:19:05 -0500
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Subject: RE: Differentiated Routing, not only plain rambo-SPF
Date: Mon, 22 Jul 2002 16:19:04 -0500
Message-ID: <A29143A40AEAE3459FF829D1DD4331613C24F3@KC-MAIL2.kc.umkc.edu>
Thread-Topic: Differentiated Routing, not only plain rambo-SPF
Thread-Index: AcIxomqvZORnaMu1SlW/dN1vQw4RiQAItVPP
From: "Ayyasamy, Senthilkumar  (UMKC-Student)" <saq66@umkc.edu>
To: "Hummel Heinrich" <Heinrich.Hummel@icn.siemens.de>,
        <routing-discussion@ietf.org>
X-OriginalArrivalTime: 22 Jul 2002 21:19:05.0271 (UTC) FILETIME=[68950470:01C231C5]
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from base64 to 8bit by optimus.ietf.org id RAA01620
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

>Differentiated Routing would certainly stress Differentiated Services 
>and I think that the reason why it hasn't been
>introduced together with DiffServ is that IETF folks are only 
>seeing the advantages of SPF routing but not its 
>disadvantages, i.e. were therefore blocked.
I have studied something reverse in some cases
(with particular reference to premium service in diffserv)
 
An excerpt from a recent infcomm paper,
* Originally, the DiffServ scheme is decoupled from IP routing 
intentionally, meaning that all traffic 
between each source-destination pair follows the same path no 
matter which service class it belongs to and DiffServ itself has no
effect on IP routing decisions. Due to premium trafficâs high priority
this could lead to some problems for the low-priority traffic when the
 volume of premium class traffic is high. That is to say,without taking 
routing into consideration, the premium class traffic imposes very 
negative influences on other classes of traffic, especially when the 
network is highly loaded.*
The authors modify SPF algo for avoiding problems in diffserv networks.
 
http://www.ieee-infocom.org/2002/papers/675.pdf
Though the paper tries to find solution for something which have been
done already (inter-class effects are discussed in the orginal EF 
RFC itself,) the authors have come up with good algebra and routing
problem analysis in diffserv networks which will help further research
in coupling IP routing and diffserv.
 
>So my initial question is: why can't we go after Differentiated Routing ?!
>And also: why can't we go after better routing paradigmas (with respective >algorithms) than SPF-routing ?!
 
As said, recent interest is using OSPF metric optimization for network planning.
People tweak around OSPF weights for managing their network. Hence there 
should be a strong reason for moving from the present routing paradigm. I am 
wary of many stateful routing schemes deployment.
 
 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Mon Jul 22 17:37:14 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA12024
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:37:14 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA04332;
	Mon, 22 Jul 2002 17:24:03 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id QAA04573
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 16:24:52 -0400 (EDT)
Received: from portal.gmu.edu (portalknot.gmu.edu [129.174.0.8])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA05790
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 16:23:50 -0400 (EDT)
From: svenkata@gmu.edu
Received: from gmu.edu (mail01.gmu.edu [129.174.0.111])
	by portal.gmu.edu (8.8.8/8.8.8) with ESMTP id QAA22211;
	Mon, 22 Jul 2002 16:24:47 -0400 (EDT)
To: fred@cisco.com
Cc: routing-discussion@ietf.org
Reply-To: svenkata@mail.gmu.edu
Message-ID: <b96485b9316a.b9316ab96485@gmu.edu>
Date: Mon, 22 Jul 2002 16:24:47 -0400
X-Mailer: Netscape Webmail
MIME-Version: 1.0
Content-Language: en
Subject: End-to-End QoS Routing
X-Accept-Language: en
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Fred,

  You completely misunderstood my good intensions.

-> >PS: I am speaking as an individual.
-> 
-> This is an IETF list. We are all individuals in this forum. 
-> Frankly, if 
-> your opinion would be different if you posted from a different email 
-> address, it's not actually your opinion, and I'm not particularly 
-> interested in it.

  As an young individual I don't want to convince you about
  my good intensions. Please read on...

-> I guess my question is "please state the problem".
-> What is the problem you want to solve, that your proposal might be a 
-> solution to?

  Hummel question is different. Both of us provided the
  same answer to Hummel question.

  My question is nothing to do with Diff-Routing. It's about
  QoS Routing w.r.t End-End behavior. I am not at all 
  asking to route at the circuit level.

  Question:
  All the current QoS routing (MPLS-TE) is limited to 
  an Area (a small fraction in the Internet). Using all 
  these QoS routing extensions are confined to these small
  areas will not buy us (an end users) anything in the 
  long run (end-to-end Internet perspective). Do you agree?

  If not,
  "How can we deliver real-time services (differentiated)
   with some end-to-end guarantee over the Internet?"

  Possible Solution(s):
  1. Hierarchical QoS Routing
  ---------------------------
  This is already an on-going discussion in IETF & IRTF:

  "Do you think that there is no such need for
   Hierarchical QoS Routing in the Internet (to provide
   differentiated service end-to-end)?" 
  
  2. Other Possible Solutions
  ---------------------------
  If we worry about scalability & aggregation of
  hierarchical QoS routing, we can search for other
  possibilities. For example, "LSP stitching" from
  source AS till destination AS. This does not 
  require any enhancements to IGPs but requires
  some way of communications of the QoS parameters
  across EGP boundaries.

  Regardless of Routing possibilities, signaling 
  protocols must provide mechanisms for accross AS 
  signaling. That is under NSIS study. Am I right?

  Again, to be specific, what are the different 
  alternatives for end-to-end QoS routing? Is 
  there any requirement to "extend" QoS routing
  beyond AS boundaries? 

  I think there is definite need (based on transport 
  area WG efforts to provide real-time services over 
  Internet). Do you agree?

  PS: Fred, I request you not to frame any opinions
      based on from address. :) SMTP doesn't validate 
      sender's address. :-)

--
Venkata.



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@ns.ietf.org  Mon Jul 22 17:37:14 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA12030
	for <routing-discussion-archive@odin.ietf.org>; Mon, 22 Jul 2002 17:37:14 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id RAA11834
	for routing-discussion-archive@odin.ietf.org; Mon, 22 Jul 2002 17:38:15 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA04332;
	Mon, 22 Jul 2002 17:24:03 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id QAA04573
	for <routing-discussion@ns.ietf.org>; Mon, 22 Jul 2002 16:24:52 -0400 (EDT)
Received: from portal.gmu.edu (portalknot.gmu.edu [129.174.0.8])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA05790
	for <routing-discussion@ietf.org>; Mon, 22 Jul 2002 16:23:50 -0400 (EDT)
From: svenkata@gmu.edu
Received: from gmu.edu (mail01.gmu.edu [129.174.0.111])
	by portal.gmu.edu (8.8.8/8.8.8) with ESMTP id QAA22211;
	Mon, 22 Jul 2002 16:24:47 -0400 (EDT)
To: fred@cisco.com
Cc: routing-discussion@ietf.org
Reply-To: svenkata@mail.gmu.edu
Message-ID: <b96485b9316a.b9316ab96485@gmu.edu>
Date: Mon, 22 Jul 2002 16:24:47 -0400
X-Mailer: Netscape Webmail
MIME-Version: 1.0
Content-Language: en
Subject: End-to-End QoS Routing
X-Accept-Language: en
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Fred,

  You completely misunderstood my good intensions.

-> >PS: I am speaking as an individual.
-> 
-> This is an IETF list. We are all individuals in this forum. 
-> Frankly, if 
-> your opinion would be different if you posted from a different email 
-> address, it's not actually your opinion, and I'm not particularly 
-> interested in it.

  As an young individual I don't want to convince you about
  my good intensions. Please read on...

-> I guess my question is "please state the problem".
-> What is the problem you want to solve, that your proposal might be a 
-> solution to?

  Hummel question is different. Both of us provided the
  same answer to Hummel question.

  My question is nothing to do with Diff-Routing. It's about
  QoS Routing w.r.t End-End behavior. I am not at all 
  asking to route at the circuit level.

  Question:
  All the current QoS routing (MPLS-TE) is limited to 
  an Area (a small fraction in the Internet). Using all 
  these QoS routing extensions are confined to these small
  areas will not buy us (an end users) anything in the 
  long run (end-to-end Internet perspective). Do you agree?

  If not,
  "How can we deliver real-time services (differentiated)
   with some end-to-end guarantee over the Internet?"

  Possible Solution(s):
  1. Hierarchical QoS Routing
  ---------------------------
  This is already an on-going discussion in IETF & IRTF:

  "Do you think that there is no such need for
   Hierarchical QoS Routing in the Internet (to provide
   differentiated service end-to-end)?" 
  
  2. Other Possible Solutions
  ---------------------------
  If we worry about scalability & aggregation of
  hierarchical QoS routing, we can search for other
  possibilities. For example, "LSP stitching" from
  source AS till destination AS. This does not 
  require any enhancements to IGPs but requires
  some way of communications of the QoS parameters
  across EGP boundaries.

  Regardless of Routing possibilities, signaling 
  protocols must provide mechanisms for accross AS 
  signaling. That is under NSIS study. Am I right?

  Again, to be specific, what are the different 
  alternatives for end-to-end QoS routing? Is 
  there any requirement to "extend" QoS routing
  beyond AS boundaries? 

  I think there is definite need (based on transport 
  area WG efforts to provide real-time services over 
  Internet). Do you agree?

  PS: Fred, I request you not to frame any opinions
      based on from address. :) SMTP doesn't validate 
      sender's address. :-)

--
Venkata.



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Tue Jul 23 09:25:56 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA10979
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 09:25:56 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id JAA09254;
	Tue, 23 Jul 2002 09:19:24 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id JAA09224
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 09:19:17 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA10623
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 09:18:15 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id PAA04565;
	Tue, 23 Jul 2002 15:18:57 +0200 (MET DST)
Received: from mchh247e.demchh201e.icn.siemens.de ([139.21.200.57])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id PAA24028;
	Tue, 23 Jul 2002 15:19:14 +0200 (MET DST)
Received: by MCHH247E with Internet Mail Service (5.5.2653.19)
	id <PPFDY8DS>; Tue, 23 Jul 2002 15:19:11 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B91D@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Naidu, Venkata'" <Venkata.Naidu@Marconi.com>,
        "'Fred Baker'"
	 <fred@cisco.com>,
        "'svenkata@gmu.edu'" <svenkata@gmu.edu>
Cc: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Tue, 23 Jul 2002 15:19:09 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hi all debaters,
Thanks for this echo on my initial email yesterday evening, European time.
May I conclude that there is indeed a lot of interest in DiffRout  (if nsis exists as a WG, but
depends on Differentiated Routing (DiffRout), then we should have a diffrout WG, too:-)

Certainly, the basic idea of DiffRout is to use the DSCP-Bits, or even all 8 TOS-bits, for addressing different
Routing Tables for looking up the next hop link (you can as well have one common Routing Table but search for the right entry while using destination address AND DSCP as search keys - this is up to the programmer).
 
The big question however is, what shall the Routing Tables, inside of multiple routers and accessed by the same DSCP, represent?
My thinking is, they should implement a particular (logical) road system - next to other road systems on top of a given
physical network infrastructure.

Some of these road systems may have overlapping links. Some of them
must not have overlapping links. Some may serve different users or different services. Let me indulge in  analogies:

A road system for pedestrians comprises side-walks, sky-walks, streets, country roads, state roads, but not highways and express ways.
A road systems for cars comprises streets, country roads, state roads, highways, express ways, conditionally toll roads, but definitively no side walks and no sky walks (only in Hollywood movies).
A road system towards downtown might also have an associated overflow road system, which is a generalized form of a detour.
And even the overflow road system may have its overflow road system...
You see, some road system may even be defined based on/dependent on the existence of other road systems.

Traffic towards downtown using DSCP X may, at the rim of downtown, be redirected, by also changing DSCP X to DSCP X-overflow, if momentarily and locally available traffic load measurement data do advise so.  

We may consider a couple of  overflow road systems whereby, e.g. the forth one, is determined by disregarding the roads of the first three road systems. These four road systems may be used alternatively so that traffic is well balanced, i.e. so that overflow situations will not occur in the first place.And yet all packets of any given micro flow would use one of these four DSCPs, so that they would all take the same route.

Or consider a road system (=detour route)which is  just a single route tunnel from one road system to the other. And if a detour route between San Francisco and L.A. is associated with DSCP=5, no one would bother if another detour route road system between N.Y. and Baltimore were associated with the same DSCP=5.

We may even allocate bandwidth to some road system (and/or to its center )and may determine its links based on 
sufficiently available bandwidth (road system admission control !).

Also: If we LOCK e.g.the pedestrian traffic onto/into one road system we keep this traffic off the other road system
(and vice versa).

In summary: this is the way I look at it. And the goal should be  to develop solutions which implements such road systems:-)
What is your opinion?

Heinrich


 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Tue Jul 23 09:25:56 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA10988
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 09:25:56 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id JAA09654
	for routing-discussion-archive@odin.ietf.org; Tue, 23 Jul 2002 09:26:57 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id JAA09254;
	Tue, 23 Jul 2002 09:19:24 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id JAA09224
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 09:19:17 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA10623
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 09:18:15 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id PAA04565;
	Tue, 23 Jul 2002 15:18:57 +0200 (MET DST)
Received: from mchh247e.demchh201e.icn.siemens.de ([139.21.200.57])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id PAA24028;
	Tue, 23 Jul 2002 15:19:14 +0200 (MET DST)
Received: by MCHH247E with Internet Mail Service (5.5.2653.19)
	id <PPFDY8DS>; Tue, 23 Jul 2002 15:19:11 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B91D@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Naidu, Venkata'" <Venkata.Naidu@Marconi.com>,
        "'Fred Baker'"
	 <fred@cisco.com>,
        "'svenkata@gmu.edu'" <svenkata@gmu.edu>
Cc: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Tue, 23 Jul 2002 15:19:09 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hi all debaters,
Thanks for this echo on my initial email yesterday evening, European time.
May I conclude that there is indeed a lot of interest in DiffRout  (if nsis exists as a WG, but
depends on Differentiated Routing (DiffRout), then we should have a diffrout WG, too:-)

Certainly, the basic idea of DiffRout is to use the DSCP-Bits, or even all 8 TOS-bits, for addressing different
Routing Tables for looking up the next hop link (you can as well have one common Routing Table but search for the right entry while using destination address AND DSCP as search keys - this is up to the programmer).
 
The big question however is, what shall the Routing Tables, inside of multiple routers and accessed by the same DSCP, represent?
My thinking is, they should implement a particular (logical) road system - next to other road systems on top of a given
physical network infrastructure.

Some of these road systems may have overlapping links. Some of them
must not have overlapping links. Some may serve different users or different services. Let me indulge in  analogies:

A road system for pedestrians comprises side-walks, sky-walks, streets, country roads, state roads, but not highways and express ways.
A road systems for cars comprises streets, country roads, state roads, highways, express ways, conditionally toll roads, but definitively no side walks and no sky walks (only in Hollywood movies).
A road system towards downtown might also have an associated overflow road system, which is a generalized form of a detour.
And even the overflow road system may have its overflow road system...
You see, some road system may even be defined based on/dependent on the existence of other road systems.

Traffic towards downtown using DSCP X may, at the rim of downtown, be redirected, by also changing DSCP X to DSCP X-overflow, if momentarily and locally available traffic load measurement data do advise so.  

We may consider a couple of  overflow road systems whereby, e.g. the forth one, is determined by disregarding the roads of the first three road systems. These four road systems may be used alternatively so that traffic is well balanced, i.e. so that overflow situations will not occur in the first place.And yet all packets of any given micro flow would use one of these four DSCPs, so that they would all take the same route.

Or consider a road system (=detour route)which is  just a single route tunnel from one road system to the other. And if a detour route between San Francisco and L.A. is associated with DSCP=5, no one would bother if another detour route road system between N.Y. and Baltimore were associated with the same DSCP=5.

We may even allocate bandwidth to some road system (and/or to its center )and may determine its links based on 
sufficiently available bandwidth (road system admission control !).

Also: If we LOCK e.g.the pedestrian traffic onto/into one road system we keep this traffic off the other road system
(and vice versa).

In summary: this is the way I look at it. And the goal should be  to develop solutions which implements such road systems:-)
What is your opinion?

Heinrich


 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Tue Jul 23 10:18:13 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA13144
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 10:18:13 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA12741;
	Tue, 23 Jul 2002 10:11:52 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA12679
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 10:11:49 -0400 (EDT)
Received: from tms001bb.han.telia.se (tms001bb.han.telia.se [131.115.230.132])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA12904
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 10:10:48 -0400 (EDT)
From: Anders.P.Bergsten@telia.se
Received: from TMS041MB.tcad.telia.se ([131.115.230.167]) by tms001bb.han.telia.se with Microsoft SMTPSVC(5.0.2195.2966);
	 Tue, 23 Jul 2002 16:11:39 +0200
Content-Class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Subject: RE: Differentiated Routing, not only plain rambo-SPF
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
Date: Tue, 23 Jul 2002 16:11:38 +0200
Message-ID: <1BF8A5964E4EFA41A094B835FCA7129110CC86@TMS041MB.tcad.telia.se>
Thread-Topic: Differentiated Routing, not only plain rambo-SPF
Thread-Index: AcIyS5pPjHBfIhRaQrG3ORoYrk002gAASP+A
To: <Heinrich.Hummel@icn.siemens.de>, <Venkata.Naidu@marconi.com>,
        <fred@cisco.com>, <svenkata@gmu.edu>
Cc: <routing-discussion@ietf.org>
X-OriginalArrivalTime: 23 Jul 2002 14:11:39.0110 (UTC) FILETIME=[DCB12860:01C23252]
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id KAA12680
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

Heinrich, 

you have touch a subject that I have been thinking on for awhile. It is definately of interest to explore such a system - I see directly some benefits of doing what you are talking about. However, I also see costs for implementing this, which demands very careful considerations before implementation/standardisation takes place. 

To answer Fred's question on why somebody would like to do such a thing: 
1) it allows for an ISP to have different downstream operators for different services (read quality). The current routing system does not (if I understand routing correctly) allow for this. 

I do not see the need for that short term. It is very interesting to read Clarks design philosophy paper though, and think about section 8, where he talks about service flexibility and that IP allow us to build network with vast different properties/services. Despite this understanding, I see no method by which we as ISPs can technically express the different goals that we have when building networks (be it cost or quality). This may lead to the conclusion that something is missing in the control infrastructure of IP, I don't know if this is the case.

Anyway, if all ISPs optimise for the same thing we clearly have no problem here. But if this is not the case then we may need something new in awhile. And, if Diffserv is to be the way by which we technically express different goals (quality) then it may be wise to re-explore the area of TOS/DSCP routing. But once again, I don't see this as a short term need.

2) it allows for dividing a physical network into logical parts as Heinrich pointed out. This can be used for example for load balancing and it also gives me possibilities to construct some new services that have attractive properties for me. More information on that may be given outside this list. 


I would not necessarily couple "Diffrouting" tightly with DSCP though, since DSCP (at least in my mind) controls queuing and you want to control the logical seperation of topology (ie to control routing). Some may want to do both, which leads to the conclusion that you would need an AND operation between the control of WHERE to send the packets and HOW to forward them. At least I think you would want to allow for that.

Anders

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Tue Jul 23 11:09:35 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA13149
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 10:18:14 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id KAA13130
	for routing-discussion-archive@odin.ietf.org; Tue, 23 Jul 2002 10:19:13 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA12741;
	Tue, 23 Jul 2002 10:11:52 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA12679
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 10:11:49 -0400 (EDT)
Received: from tms001bb.han.telia.se (tms001bb.han.telia.se [131.115.230.132])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA12904
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 10:10:48 -0400 (EDT)
From: Anders.P.Bergsten@telia.se
Received: from TMS041MB.tcad.telia.se ([131.115.230.167]) by tms001bb.han.telia.se with Microsoft SMTPSVC(5.0.2195.2966);
	 Tue, 23 Jul 2002 16:11:39 +0200
Content-Class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Subject: RE: Differentiated Routing, not only plain rambo-SPF
X-MimeOLE: Produced By Microsoft Exchange V6.0.5762.3
Date: Tue, 23 Jul 2002 16:11:38 +0200
Message-ID: <1BF8A5964E4EFA41A094B835FCA7129110CC86@TMS041MB.tcad.telia.se>
Thread-Topic: Differentiated Routing, not only plain rambo-SPF
Thread-Index: AcIyS5pPjHBfIhRaQrG3ORoYrk002gAASP+A
To: <Heinrich.Hummel@icn.siemens.de>, <Venkata.Naidu@marconi.com>,
        <fred@cisco.com>, <svenkata@gmu.edu>
Cc: <routing-discussion@ietf.org>
X-OriginalArrivalTime: 23 Jul 2002 14:11:39.0110 (UTC) FILETIME=[DCB12860:01C23252]
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id KAA12680
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

Heinrich, 

you have touch a subject that I have been thinking on for awhile. It is definately of interest to explore such a system - I see directly some benefits of doing what you are talking about. However, I also see costs for implementing this, which demands very careful considerations before implementation/standardisation takes place. 

To answer Fred's question on why somebody would like to do such a thing: 
1) it allows for an ISP to have different downstream operators for different services (read quality). The current routing system does not (if I understand routing correctly) allow for this. 

I do not see the need for that short term. It is very interesting to read Clarks design philosophy paper though, and think about section 8, where he talks about service flexibility and that IP allow us to build network with vast different properties/services. Despite this understanding, I see no method by which we as ISPs can technically express the different goals that we have when building networks (be it cost or quality). This may lead to the conclusion that something is missing in the control infrastructure of IP, I don't know if this is the case.

Anyway, if all ISPs optimise for the same thing we clearly have no problem here. But if this is not the case then we may need something new in awhile. And, if Diffserv is to be the way by which we technically express different goals (quality) then it may be wise to re-explore the area of TOS/DSCP routing. But once again, I don't see this as a short term need.

2) it allows for dividing a physical network into logical parts as Heinrich pointed out. This can be used for example for load balancing and it also gives me possibilities to construct some new services that have attractive properties for me. More information on that may be given outside this list. 


I would not necessarily couple "Diffrouting" tightly with DSCP though, since DSCP (at least in my mind) controls queuing and you want to control the logical seperation of topology (ie to control routing). Some may want to do both, which leads to the conclusion that you would need an AND operation between the control of WHERE to send the packets and HOW to forward them. At least I think you would want to allow for that.

Anders

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Tue Jul 23 11:13:22 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA16876
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 11:13:22 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA15592;
	Tue, 23 Jul 2002 10:56:17 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA15564
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 10:56:15 -0400 (EDT)
Received: from gorilla.mchh.siemens.de (gorilla.mchh.siemens.de [194.138.158.18])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA15376
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 10:55:13 -0400 (EDT)
Received: from blues.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.227])
	by gorilla.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id QAA17959;
	Tue, 23 Jul 2002 16:56:06 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by blues.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id QAA04851;
	Tue, 23 Jul 2002 16:56:07 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2653.19)
	id <PPBZKA89>; Tue, 23 Jul 2002 16:56:05 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B91F@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Anders.P.Bergsten@telia.se'" <Anders.P.Bergsten@telia.se>,
        Venkata.Naidu@marconi.com, fred@cisco.com, svenkata@gmu.edu
Cc: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Tue, 23 Jul 2002 16:55:56 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id KAA15565
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

Anders,
One more analogy from our daily road traffic:
DiffServ is what motor-cyclists enjoy at a red traffic light: they can jump the queue of waiting cars.
And this is really not too fancy compared with  all the other mechanisms for managing the road traffic.
DiffServ and DiffRout : How to share the codepoint ? 
DSCP for Gold service = DRCP for using all roads (i.e. incl. express ways, toll roads)
DSCP for Silver service = DRCP for using all roads except...
DSCP for Bronze service = DRCP for using country roads only unless some better road is needed for reaching the destination.

The lesser service traffic is kept off from the best roads of the better services' road systems.
I.e. service differentiation is stressed, both for the better and for the worse.  

Of course life would be easier if we wouldn't have to double the interpretation of the DSCP-bits. But
a) no ISP supports 64 different SLAs
b) see above, some of the road systems may be in compliance/in support of service differentiation.

E.g. a particular bunch of DSCPs may share some DiffServ meaning, although they addresses different road systems.

Maybe the flow label in IPv6 could be used, too. But I think these are second-degree details.

Heinrich

-----Ursprüngliche Nachricht-----
Von: Anders.P.Bergsten@telia.se [mailto:Anders.P.Bergsten@telia.se]
Gesendet: Dienstag, 23. Juli 2002 16:12
An: Heinrich.Hummel@icn.siemens.de; Venkata.Naidu@marconi.com;
fred@cisco.com; svenkata@gmu.edu
Cc: routing-discussion@ietf.org
Betreff: RE: Differentiated Routing, not only plain rambo-SPF


Heinrich, 

you have touch a subject that I have been thinking on for awhile. It is definately of interest to explore such a system - I see directly some benefits of doing what you are talking about. However, I also see costs for implementing this, which demands very careful considerations before implementation/standardisation takes place. 

To answer Fred's question on why somebody would like to do such a thing: 
1) it allows for an ISP to have different downstream operators for different services (read quality). The current routing system does not (if I understand routing correctly) allow for this. 

I do not see the need for that short term. It is very interesting to read Clarks design philosophy paper though, and think about section 8, where he talks about service flexibility and that IP allow us to build network with vast different properties/services. Despite this understanding, I see no method by which we as ISPs can technically express the different goals that we have when building networks (be it cost or quality). This may lead to the conclusion that something is missing in the control infrastructure of IP, I don't know if this is the case.

Anyway, if all ISPs optimise for the same thing we clearly have no problem here. But if this is not the case then we may need something new in awhile. And, if Diffserv is to be the way by which we technically express different goals (quality) then it may be wise to re-explore the area of TOS/DSCP routing. But once again, I don't see this as a short term need.

2) it allows for dividing a physical network into logical parts as Heinrich pointed out. This can be used for example for load balancing and it also gives me possibilities to construct some new services that have attractive properties for me. More information on that may be given outside this list. 


I would not necessarily couple "Diffrouting" tightly with DSCP though, since DSCP (at least in my mind) controls queuing and you want to control the logical seperation of topology (ie to control routing). Some may want to do both, which leads to the conclusion that you would need an AND operation between the control of WHERE to send the packets and HOW to forward them. At least I think you would want to allow for that.

Anders

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Tue Jul 23 12:09:34 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA16882
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 11:13:23 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id LAA17672
	for routing-discussion-archive@odin.ietf.org; Tue, 23 Jul 2002 11:14:23 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA15592;
	Tue, 23 Jul 2002 10:56:17 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id KAA15564
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 10:56:15 -0400 (EDT)
Received: from gorilla.mchh.siemens.de (gorilla.mchh.siemens.de [194.138.158.18])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA15376
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 10:55:13 -0400 (EDT)
Received: from blues.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.227])
	by gorilla.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id QAA17959;
	Tue, 23 Jul 2002 16:56:06 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by blues.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id QAA04851;
	Tue, 23 Jul 2002 16:56:07 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2653.19)
	id <PPBZKA89>; Tue, 23 Jul 2002 16:56:05 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B91F@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Anders.P.Bergsten@telia.se'" <Anders.P.Bergsten@telia.se>,
        Venkata.Naidu@marconi.com, fred@cisco.com, svenkata@gmu.edu
Cc: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Tue, 23 Jul 2002 16:55:56 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by optimus.ietf.org id KAA15565
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 8bit

Anders,
One more analogy from our daily road traffic:
DiffServ is what motor-cyclists enjoy at a red traffic light: they can jump the queue of waiting cars.
And this is really not too fancy compared with  all the other mechanisms for managing the road traffic.
DiffServ and DiffRout : How to share the codepoint ? 
DSCP for Gold service = DRCP for using all roads (i.e. incl. express ways, toll roads)
DSCP for Silver service = DRCP for using all roads except...
DSCP for Bronze service = DRCP for using country roads only unless some better road is needed for reaching the destination.

The lesser service traffic is kept off from the best roads of the better services' road systems.
I.e. service differentiation is stressed, both for the better and for the worse.  

Of course life would be easier if we wouldn't have to double the interpretation of the DSCP-bits. But
a) no ISP supports 64 different SLAs
b) see above, some of the road systems may be in compliance/in support of service differentiation.

E.g. a particular bunch of DSCPs may share some DiffServ meaning, although they addresses different road systems.

Maybe the flow label in IPv6 could be used, too. But I think these are second-degree details.

Heinrich

-----Ursprüngliche Nachricht-----
Von: Anders.P.Bergsten@telia.se [mailto:Anders.P.Bergsten@telia.se]
Gesendet: Dienstag, 23. Juli 2002 16:12
An: Heinrich.Hummel@icn.siemens.de; Venkata.Naidu@marconi.com;
fred@cisco.com; svenkata@gmu.edu
Cc: routing-discussion@ietf.org
Betreff: RE: Differentiated Routing, not only plain rambo-SPF


Heinrich, 

you have touch a subject that I have been thinking on for awhile. It is definately of interest to explore such a system - I see directly some benefits of doing what you are talking about. However, I also see costs for implementing this, which demands very careful considerations before implementation/standardisation takes place. 

To answer Fred's question on why somebody would like to do such a thing: 
1) it allows for an ISP to have different downstream operators for different services (read quality). The current routing system does not (if I understand routing correctly) allow for this. 

I do not see the need for that short term. It is very interesting to read Clarks design philosophy paper though, and think about section 8, where he talks about service flexibility and that IP allow us to build network with vast different properties/services. Despite this understanding, I see no method by which we as ISPs can technically express the different goals that we have when building networks (be it cost or quality). This may lead to the conclusion that something is missing in the control infrastructure of IP, I don't know if this is the case.

Anyway, if all ISPs optimise for the same thing we clearly have no problem here. But if this is not the case then we may need something new in awhile. And, if Diffserv is to be the way by which we technically express different goals (quality) then it may be wise to re-explore the area of TOS/DSCP routing. But once again, I don't see this as a short term need.

2) it allows for dividing a physical network into logical parts as Heinrich pointed out. This can be used for example for load balancing and it also gives me possibilities to construct some new services that have attractive properties for me. More information on that may be given outside this list. 


I would not necessarily couple "Diffrouting" tightly with DSCP though, since DSCP (at least in my mind) controls queuing and you want to control the logical seperation of topology (ie to control routing). Some may want to do both, which leads to the conclusion that you would need an AND operation between the control of WHERE to send the packets and HOW to forward them. At least I think you would want to allow for that.

Anders

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Tue Jul 23 13:31:47 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA25357
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 13:31:47 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA29844;
	Tue, 23 Jul 2002 13:25:05 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA29816
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 13:25:03 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA24958
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 13:24:00 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6NHOThJ025808;
	Tue, 23 Jul 2002 10:24:30 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020723102319.00a8a930@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Tue, 23 Jul 2002 10:24:27 -0700
To: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
From: Fred Baker <fred@cisco.com>
Subject: Re: AW: Differentiated Routing, not only plain rambo-SPF
Cc: routing-discussion@ietf.org
In-Reply-To: <EF8E39AA846CD411BB9C00508B951F510514B91D@MCHH267E>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

At 03:19 PM 7/23/2002 +0200, Hummel Heinrich wrote:
>The big question however is, what shall the Routing Tables, inside of 
>multiple routers and accessed by the same DSCP, represent?

no. The big question is "what problem are you trying to solve?" If it is a 
lack of roads/circuits/channels in IP routing, I'm not convinced that is a 
problem.


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Tue Jul 23 13:31:48 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA25365
	for <routing-discussion-archive@odin.ietf.org>; Tue, 23 Jul 2002 13:31:48 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id NAA01211
	for routing-discussion-archive@odin.ietf.org; Tue, 23 Jul 2002 13:32:49 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA29844;
	Tue, 23 Jul 2002 13:25:05 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id NAA29816
	for <routing-discussion@optimus.ietf.org>; Tue, 23 Jul 2002 13:25:03 -0400 (EDT)
Received: from sj-msg-core-2.cisco.com (sj-msg-core-2.cisco.com [171.69.24.11])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA24958
	for <routing-discussion@ietf.org>; Tue, 23 Jul 2002 13:24:00 -0400 (EDT)
Received: from FRED-W2K6.cisco.com (stealth-10-32-253-236.cisco.com [10.32.253.236])
	by sj-msg-core-2.cisco.com (8.12.2/8.12.2) with SMTP id g6NHOThJ025808;
	Tue, 23 Jul 2002 10:24:30 -0700 (PDT)
Message-Id: <5.1.1.6.2.20020723102319.00a8a930@mira-sjcm-4.cisco.com>
X-Sender: fred@mira-sjcm-4.cisco.com
X-Mailer: QUALCOMM Windows Eudora Version 5.1.1
Date: Tue, 23 Jul 2002 10:24:27 -0700
To: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
From: Fred Baker <fred@cisco.com>
Subject: Re: AW: Differentiated Routing, not only plain rambo-SPF
Cc: routing-discussion@ietf.org
In-Reply-To: <EF8E39AA846CD411BB9C00508B951F510514B91D@MCHH267E>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

At 03:19 PM 7/23/2002 +0200, Hummel Heinrich wrote:
>The big question however is, what shall the Routing Tables, inside of 
>multiple routers and accessed by the same DSCP, represent?

no. The big question is "what problem are you trying to solve?" If it is a 
lack of roads/circuits/channels in IP routing, I'm not convinced that is a 
problem.


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From daemon@optimus.ietf.org  Wed Jul 24 08:16:39 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA05079
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 08:16:38 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id IAA18590
	for routing-discussion-archive@odin.ietf.org; Wed, 24 Jul 2002 08:17:41 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id IAA17402;
	Wed, 24 Jul 2002 08:11:04 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id IAA17337
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 08:10:58 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA02953
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 08:09:54 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id KAA06807;
	Wed, 24 Jul 2002 10:48:32 +0200 (MET DST)
Received: from mchh247e.demchh201e.icn.siemens.de ([139.21.200.57])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id KAA06268;
	Wed, 24 Jul 2002 10:48:50 +0200 (MET DST)
Received: by MCHH247E with Internet Mail Service (5.5.2653.19)
	id <PPFDZRFP>; Wed, 24 Jul 2002 10:48:48 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B923@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Fred Baker'" <fred@cisco.com>
Cc: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Wed, 24 Jul 2002 10:48:48 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Fred,
Don't over-esteem my phrasing. Maybe I should have written:
The next question ... instead of The big question.... 

I started this thread with the following two sentences:
In normal, state-less IP routing we cannot do QoS-Routing, nor sneak-around the congestion-area routing, nor traffic balanced routing. Instead, all of this is left to MPLS, where multiple states and multiple LSPs and multiple signalling is required.
With other words:
I try to enable QoS-Routing, "sneak-around the congestion-area" -Routing, traffic balanced -Routing. 
Without MPLS and without RSVP.Without circuits and without channels. I also pointed out that you may get for free
multicast delivery channels with voice/video quality. 

From the response I learnt that others have already tried the same, see:
http://www.ieee-infocom.org/2002/papers/675.pdf
and have even chosen the same name = Differentiated Routing. For some reasons no one knows or can remember study on
Differentiated Routing was discontinued  and different prioritization became the seul topic of DiffServ.

Furthermore, it was not me who really started this discussion. Instead it was the IESG, in Salt Lake City (I think),
that encouraged discussion on Equal-Cost Multipath and I guess, as a consequence, this routing-discussion mailing list
has been installed. But from a former admiror of ECMP I turned to opposing it. Hashing traffic according to source/destination is not the best thing. I can hardly imagine that a traffic policeman who has to balance the oncoming
traffic by sending a third to the left, a third to the middle, and a third to the right would ever get the idea 
to check the license plates of the cars, and e.g. would send those from Florida,Georgia,Nevada,.. to the left, 
those from California,Wyoming, Kansas,... to the middle, those from New York,Texas,... to the right.
What would garantee him that the chosen hashing would yield the right proportioning ? 
What if he fails ? Rearranging the three sets of US-states ? But who knows which cars may arrive later on?
I think he will never find out what are the right three sets of US-states.
Believe me, this is the poorest solution ever.

Heinrich



At 03:19 PM 7/23/2002 +0200, Hummel Heinrich wrote:
>The big question however is, what shall the Routing Tables, inside of 
>multiple routers and accessed by the same DSCP, represent?

no. The big question is "what problem are you trying to solve?" If it is a 
lack of roads/circuits/channels in IP routing, I'm not convinced that is a 
problem.



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Wed Jul 24 09:04:06 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA05071
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 08:16:38 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id IAA17402;
	Wed, 24 Jul 2002 08:11:04 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id IAA17337
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 08:10:58 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA02953
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 08:09:54 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id KAA06807;
	Wed, 24 Jul 2002 10:48:32 +0200 (MET DST)
Received: from mchh247e.demchh201e.icn.siemens.de ([139.21.200.57])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id KAA06268;
	Wed, 24 Jul 2002 10:48:50 +0200 (MET DST)
Received: by MCHH247E with Internet Mail Service (5.5.2653.19)
	id <PPFDZRFP>; Wed, 24 Jul 2002 10:48:48 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B923@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'Fred Baker'" <fred@cisco.com>
Cc: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Wed, 24 Jul 2002 10:48:48 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Fred,
Don't over-esteem my phrasing. Maybe I should have written:
The next question ... instead of The big question.... 

I started this thread with the following two sentences:
In normal, state-less IP routing we cannot do QoS-Routing, nor sneak-around the congestion-area routing, nor traffic balanced routing. Instead, all of this is left to MPLS, where multiple states and multiple LSPs and multiple signalling is required.
With other words:
I try to enable QoS-Routing, "sneak-around the congestion-area" -Routing, traffic balanced -Routing. 
Without MPLS and without RSVP.Without circuits and without channels. I also pointed out that you may get for free
multicast delivery channels with voice/video quality. 

From the response I learnt that others have already tried the same, see:
http://www.ieee-infocom.org/2002/papers/675.pdf
and have even chosen the same name = Differentiated Routing. For some reasons no one knows or can remember study on
Differentiated Routing was discontinued  and different prioritization became the seul topic of DiffServ.

Furthermore, it was not me who really started this discussion. Instead it was the IESG, in Salt Lake City (I think),
that encouraged discussion on Equal-Cost Multipath and I guess, as a consequence, this routing-discussion mailing list
has been installed. But from a former admiror of ECMP I turned to opposing it. Hashing traffic according to source/destination is not the best thing. I can hardly imagine that a traffic policeman who has to balance the oncoming
traffic by sending a third to the left, a third to the middle, and a third to the right would ever get the idea 
to check the license plates of the cars, and e.g. would send those from Florida,Georgia,Nevada,.. to the left, 
those from California,Wyoming, Kansas,... to the middle, those from New York,Texas,... to the right.
What would garantee him that the chosen hashing would yield the right proportioning ? 
What if he fails ? Rearranging the three sets of US-states ? But who knows which cars may arrive later on?
I think he will never find out what are the right three sets of US-states.
Believe me, this is the poorest solution ever.

Heinrich



At 03:19 PM 7/23/2002 +0200, Hummel Heinrich wrote:
>The big question however is, what shall the Routing Tables, inside of 
>multiple routers and accessed by the same DSCP, represent?

no. The big question is "what problem are you trying to solve?" If it is a 
lack of roads/circuits/channels in IP routing, I'm not convinced that is a 
problem.



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From routing-discussion-admin@ietf.org  Wed Jul 24 12:43:14 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA23457
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 12:43:14 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA09232;
	Wed, 24 Jul 2002 12:36:12 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA09190
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 12:36:10 -0400 (EDT)
Received: from ginger.lcs.mit.edu (ginger.lcs.mit.edu [18.26.0.82])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA23232
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 12:35:06 -0400 (EDT)
Received: (from jnc@localhost)
	by ginger.lcs.mit.edu (8.9.1/8.9.1) id MAA15741;
	Wed, 24 Jul 2002 12:35:03 -0400
Date: Wed, 24 Jul 2002 12:35:03 -0400
From: "J. Noel Chiappa" <jnc@ginger.lcs.mit.edu>
Message-Id: <200207241635.MAA15741@ginger.lcs.mit.edu>
To: routing-discussion@ietf.org
Subject: Re:  AW: Differentiated Routing, not only plain rambo-SPF
Cc: jnc@ginger.lcs.mit.edu
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

    > From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>

    > In normal, state-less IP routing

You just hit one of my hot buttons.

Normal IP routing is not stateless. It has lots of state in the network.
Moreover, that state is state such that if it is not properly maintained will
cause packets not to flow. In other words, it is also critical (in the sense
of "it's right, or the communication doesn't work") state. RIP, OSPF, IS-IS,
BGP - they all maintain state, and often include extremely complex and
powerful mechanisms to ensure that that state is perfectly synchronized -
because if it isn't synchronized perfectly, and up-to-date, things will go to
hell in a handbasket. 


The way I look at routing/forwarding, information is split between two places:
the packet header, and the switches (akak routers). A combination of the two
is necessary to determine what path the packet is going to take. The
information in the header allows the switch/router to look into the
information (state) is has, and figure out what to do next.

In this high-level model, it's completely irrelevant whether the information
in the header is:

- just a destination address (in which case the paths selected are independent
	of any kind of service)
- a destination address plus some sort of service selection (be it ToS, or
	DiffServ, or whatever)
- a path selector (think ATM or MPLS)
- a loose source route
- a complete source route

although I admit that in the last one, the information in the switches is
vestigial (simply a binding between the interfaces identifiers in the packet,
and the actual interfaces).

In any routing system, the designer has to answer the following questions:

i) what kind of services do you want the user to be able to ask for;
ii) how do they ask for it (e.g. simply with bits in each packet header, or
	what);
iii) what information does this mean you have to have in the switches, and how
	does it get there.

"Classic" IP routing has a particular set of answers to these questions - and
it does involve state in the routers.


We now return you to the previous discussion...

	Noel

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Wed Jul 24 13:34:06 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA23466
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 12:43:15 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id MAA09694
	for routing-discussion-archive@odin.ietf.org; Wed, 24 Jul 2002 12:44:18 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA09232;
	Wed, 24 Jul 2002 12:36:12 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA09190
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 12:36:10 -0400 (EDT)
Received: from ginger.lcs.mit.edu (ginger.lcs.mit.edu [18.26.0.82])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA23232
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 12:35:06 -0400 (EDT)
Received: (from jnc@localhost)
	by ginger.lcs.mit.edu (8.9.1/8.9.1) id MAA15741;
	Wed, 24 Jul 2002 12:35:03 -0400
Date: Wed, 24 Jul 2002 12:35:03 -0400
From: "J. Noel Chiappa" <jnc@ginger.lcs.mit.edu>
Message-Id: <200207241635.MAA15741@ginger.lcs.mit.edu>
To: routing-discussion@ietf.org
Subject: Re:  AW: Differentiated Routing, not only plain rambo-SPF
Cc: jnc@ginger.lcs.mit.edu
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

    > From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>

    > In normal, state-less IP routing

You just hit one of my hot buttons.

Normal IP routing is not stateless. It has lots of state in the network.
Moreover, that state is state such that if it is not properly maintained will
cause packets not to flow. In other words, it is also critical (in the sense
of "it's right, or the communication doesn't work") state. RIP, OSPF, IS-IS,
BGP - they all maintain state, and often include extremely complex and
powerful mechanisms to ensure that that state is perfectly synchronized -
because if it isn't synchronized perfectly, and up-to-date, things will go to
hell in a handbasket. 


The way I look at routing/forwarding, information is split between two places:
the packet header, and the switches (akak routers). A combination of the two
is necessary to determine what path the packet is going to take. The
information in the header allows the switch/router to look into the
information (state) is has, and figure out what to do next.

In this high-level model, it's completely irrelevant whether the information
in the header is:

- just a destination address (in which case the paths selected are independent
	of any kind of service)
- a destination address plus some sort of service selection (be it ToS, or
	DiffServ, or whatever)
- a path selector (think ATM or MPLS)
- a loose source route
- a complete source route

although I admit that in the last one, the information in the switches is
vestigial (simply a binding between the interfaces identifiers in the packet,
and the actual interfaces).

In any routing system, the designer has to answer the following questions:

i) what kind of services do you want the user to be able to ask for;
ii) how do they ask for it (e.g. simply with bits in each packet header, or
	what);
iii) what information does this mean you have to have in the switches, and how
	does it get there.

"Classic" IP routing has a particular set of answers to these questions - and
it does involve state in the routers.


We now return you to the previous discussion...

	Noel

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Wed Jul 24 14:45:13 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27303
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 14:45:13 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA15503;
	Wed, 24 Jul 2002 14:38:50 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA15472
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 14:38:48 -0400 (EDT)
Received: from mailgate.pit.comms.marconi.com (mailgate.pit.comms.marconi.com [169.144.68.6])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27086
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 14:37:45 -0400 (EDT)
Received: from mailman.pit.comms.marconi.com (mailman.pit.comms.marconi.com [169.144.2.12])
	by mailgate.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA24758;
	Wed, 24 Jul 2002 14:38:08 -0400 (EDT)
Received: from whq-msgrtr-01.pit.comms.marconi.com (whq-msgrtr-01.pit.comms.marconi.com [169.144.2.221])
	by mailman.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA26398;
	Wed, 24 Jul 2002 14:38:10 -0400 (EDT)
Received: by whq-msgrtr-01.pit.comms.marconi.com with Internet Mail Service (5.5.2650.21)
	id <3W99A444>; Wed, 24 Jul 2002 14:38:09 -0400
Message-ID: <39469E08BD83D411A3D900204840EC55763278@vie-msgusr-01.dc.fore.com>
From: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
To: "'J. Noel Chiappa'" <jnc@ginger.lcs.mit.edu>, routing-discussion@ietf.org,
        "'fred@cisco.com'" <fred@cisco.com>
Subject: RE: AW: Differentiated Routing, not only plain rambo-SPF
Date: Wed, 24 Jul 2002 14:38:05 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Noel,

-> From: J. Noel Chiappa [mailto:jnc@ginger.lcs.mit.edu]
->     > From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
-> 
->     > In normal, state-less IP routing
-> 
-> You just hit one of my hot buttons.
-> 
-> Normal IP routing is not stateless. It has lots of state in 
-> the network.
  
  Hummel is not talking about *state* in the network.
  The information (in your words: a Header) required to 
  route packets is state-less. Routing is purely based on
  *destination* address. Isn't it?
  
-> - just a destination address (in which case the paths 
-> selected are independent of any kind of service)

This is what Hummel meant by *Classical state-less IP-Routing*

  All the below *routing schemes* are considered as
  extra states (in addition to destination address)

-> - a destination address plus some sort of service selection 
-> (be it ToS, or	DiffServ, or whatever)
-> - a path selector (think ATM or MPLS)
-> - a loose source route
-> - a complete source route

  IMO, most of the confusions in the lists are because of 
  our terminology. I have been trying to convince Fred
  about the difference and what we need to achieve based
  on *Traffic Management* and *Traffic Engineering*.
  But I failed to express in the right way.

  I know...I know that Traffic "Engineering" is for
  efficient use of network resources in SPs. Where as
  Traffic "Management" is a local decision on how a
  router must forward/queue/schedule packets to achieve
  a desired CoS QoS.

  IMHO, just using DSCP bits on per-hop won't achieve 
  QoS "Globally". You have to combine TE also.
  That "QoS using TE" has been neglected by us.
  That is what I am trying to convince Fred. 
  
  If you consider IP is a superset, MPLS, QoS and TE are
  some overlapping subsets (pictorially):

       IP Superset
       +-----------------------------------+
       |     __________    __________      |
       |    /          \  /           \    |
       |   /   MPLS     \/    QoS      \   | 
       |  /             /\              \  | 
       | |             |  |             |  | 
       | |             |  |             |  |
       | |             |  |             |  |
       | |          ___|__|___          |  |
       |  \        /    \/    \        /   |
       |   \      /     /\     \      /    |
       |    \____/_____/  \_____\____/     |
       |        |                |         |    
       |        |                |         |
       |        |                |         |
       |         \      TE      /          |
       |          \            /           |
       |           \__________/            |
       |                                   |
       +-----------------------------------+

   You can just do MPLS (for fast label look-up etc).
   You can achieve QoS using MPLS (without TE)
   You can achieve better QoS using TE (with or 
   without MPLS) etc etc...

   IMHO, the intersection of all the subsets (ie., 
   (QoS using MPLS-(DiffServ-Aware-TE) will make a 
   *real-good* impact on end-to-end performance in the 
   global Internet.
   
   I am not alleging these points here. There has been
   some research & simulation in this area. Please 
   look at an old neglected paper by Raj Jain's group: 
ftp://ftp.netlab.ohio-state.edu/pub/jain/ietf/draft-bhani-mpls-te-anal-00.tx
t
   
  Simulation & analysis in a simple topology shows:
  "..in order to benefit from traffic engineering, MPLS 
   trunks should be implemented end-to-end (first router
   to last router). If some part of the network is 
   MPLS-unaware, the benefits are reduced or eliminated."

  So, are we doing MPLS-TE end-to-end to achieve desired QoS?
  If not (I am sure we are not - we are still doing MPLS-TE
  in small areas), then what are the alternatives?

--
Venkata.

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Wed Jul 24 14:45:13 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27311
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 14:45:13 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id OAA15908
	for routing-discussion-archive@odin.ietf.org; Wed, 24 Jul 2002 14:46:16 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA15503;
	Wed, 24 Jul 2002 14:38:50 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id OAA15472
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 14:38:48 -0400 (EDT)
Received: from mailgate.pit.comms.marconi.com (mailgate.pit.comms.marconi.com [169.144.68.6])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27086
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 14:37:45 -0400 (EDT)
Received: from mailman.pit.comms.marconi.com (mailman.pit.comms.marconi.com [169.144.2.12])
	by mailgate.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA24758;
	Wed, 24 Jul 2002 14:38:08 -0400 (EDT)
Received: from whq-msgrtr-01.pit.comms.marconi.com (whq-msgrtr-01.pit.comms.marconi.com [169.144.2.221])
	by mailman.pit.comms.marconi.com (8.9.3/8.9.3) with ESMTP id OAA26398;
	Wed, 24 Jul 2002 14:38:10 -0400 (EDT)
Received: by whq-msgrtr-01.pit.comms.marconi.com with Internet Mail Service (5.5.2650.21)
	id <3W99A444>; Wed, 24 Jul 2002 14:38:09 -0400
Message-ID: <39469E08BD83D411A3D900204840EC55763278@vie-msgusr-01.dc.fore.com>
From: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
To: "'J. Noel Chiappa'" <jnc@ginger.lcs.mit.edu>, routing-discussion@ietf.org,
        "'fred@cisco.com'" <fred@cisco.com>
Subject: RE: AW: Differentiated Routing, not only plain rambo-SPF
Date: Wed, 24 Jul 2002 14:38:05 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Noel,

-> From: J. Noel Chiappa [mailto:jnc@ginger.lcs.mit.edu]
->     > From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
-> 
->     > In normal, state-less IP routing
-> 
-> You just hit one of my hot buttons.
-> 
-> Normal IP routing is not stateless. It has lots of state in 
-> the network.
  
  Hummel is not talking about *state* in the network.
  The information (in your words: a Header) required to 
  route packets is state-less. Routing is purely based on
  *destination* address. Isn't it?
  
-> - just a destination address (in which case the paths 
-> selected are independent of any kind of service)

This is what Hummel meant by *Classical state-less IP-Routing*

  All the below *routing schemes* are considered as
  extra states (in addition to destination address)

-> - a destination address plus some sort of service selection 
-> (be it ToS, or	DiffServ, or whatever)
-> - a path selector (think ATM or MPLS)
-> - a loose source route
-> - a complete source route

  IMO, most of the confusions in the lists are because of 
  our terminology. I have been trying to convince Fred
  about the difference and what we need to achieve based
  on *Traffic Management* and *Traffic Engineering*.
  But I failed to express in the right way.

  I know...I know that Traffic "Engineering" is for
  efficient use of network resources in SPs. Where as
  Traffic "Management" is a local decision on how a
  router must forward/queue/schedule packets to achieve
  a desired CoS QoS.

  IMHO, just using DSCP bits on per-hop won't achieve 
  QoS "Globally". You have to combine TE also.
  That "QoS using TE" has been neglected by us.
  That is what I am trying to convince Fred. 
  
  If you consider IP is a superset, MPLS, QoS and TE are
  some overlapping subsets (pictorially):

       IP Superset
       +-----------------------------------+
       |     __________    __________      |
       |    /          \  /           \    |
       |   /   MPLS     \/    QoS      \   | 
       |  /             /\              \  | 
       | |             |  |             |  | 
       | |             |  |             |  |
       | |             |  |             |  |
       | |          ___|__|___          |  |
       |  \        /    \/    \        /   |
       |   \      /     /\     \      /    |
       |    \____/_____/  \_____\____/     |
       |        |                |         |    
       |        |                |         |
       |        |                |         |
       |         \      TE      /          |
       |          \            /           |
       |           \__________/            |
       |                                   |
       +-----------------------------------+

   You can just do MPLS (for fast label look-up etc).
   You can achieve QoS using MPLS (without TE)
   You can achieve better QoS using TE (with or 
   without MPLS) etc etc...

   IMHO, the intersection of all the subsets (ie., 
   (QoS using MPLS-(DiffServ-Aware-TE) will make a 
   *real-good* impact on end-to-end performance in the 
   global Internet.
   
   I am not alleging these points here. There has been
   some research & simulation in this area. Please 
   look at an old neglected paper by Raj Jain's group: 
ftp://ftp.netlab.ohio-state.edu/pub/jain/ietf/draft-bhani-mpls-te-anal-00.tx
t
   
  Simulation & analysis in a simple topology shows:
  "..in order to benefit from traffic engineering, MPLS 
   trunks should be implemented end-to-end (first router
   to last router). If some part of the network is 
   MPLS-unaware, the benefits are reduced or eliminated."

  So, are we doing MPLS-TE end-to-end to achieve desired QoS?
  If not (I am sure we are not - we are still doing MPLS-TE
  in small areas), then what are the alternatives?

--
Venkata.

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Wed Jul 24 15:14:31 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA28582
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 15:14:31 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA17010;
	Wed, 24 Jul 2002 15:06:55 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA16980
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 15:06:45 -0400 (EDT)
Received: from stl-smtpout-01.boeing.com (stl-smtpout-01.boeing.com [12.13.247.21])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA28248
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 15:05:35 -0400 (EDT)
Received: from stl-av-02.boeing.com ([192.76.190.7])
	by stl-smtpout-01.boeing.com (8.9.2/8.8.5-M2) with ESMTP id OAA00194;
	Wed, 24 Jul 2002 14:06:35 -0500 (CDT)
Received: from blv-hub-01.boeing.com (localhost [127.0.0.1])
	by stl-av-02.boeing.com (8.9.3/8.9.2/MBS-AV-01) with ESMTP id OAA19364;
	Wed, 24 Jul 2002 14:06:23 -0500 (CDT)
Received: from xch-phl-02p.ne.nos.boeing.com (xch-phl-02p.ne.nos.boeing.com [128.225.22.4])
	by blv-hub-01.boeing.com (8.11.3/8.11.3/MBS-LDAP-01) with ESMTP id g6OJ6X629475;
	Wed, 24 Jul 2002 12:06:33 -0700 (PDT)
Received: by xch-phl-02p.ne.nos.boeing.com with Internet Mail Service (5.5.2650.21)
	id <37GWZQFP>; Wed, 24 Jul 2002 15:06:31 -0400
Message-ID: <8D6509B44ACCEB43A96E032B9C3EA0C8078A3A43@xch-ne-02.ne.nos.boeing.com>
From: "Manfredi, Albert E" <albert.e.manfredi@boeing.com>
To: "'Naidu, Venkata'" <Venkata.Naidu@marconi.com>,
        routing-discussion@ietf.org
Subject: RE: AW: Differentiated Routing, not only plain rambo-SPF
Date: Wed, 24 Jul 2002 15:06:25 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

> -----Original Message-----
> From: Naidu, Venkata [mailto:Venkata.Naidu@marconi.com]

> -> From: J. Noel Chiappa [mailto:jnc@ginger.lcs.mit.edu]

> -> 
> -> You just hit one of my hot buttons.
> -> 
> -> Normal IP routing is not stateless. It has lots of state in 
> -> the network.
>   
>   Hummel is not talking about *state* in the network.
>   The information (in your words: a Header) required to 
>   route packets is state-less. Routing is purely based on
>   *destination* address. Isn't it?

But the routers must maintain state information, such as network topology (OSPF) or spanning tree (RIP). True, under default IP best-effort routing, routers do not need to maintain state information *per flow*. But in case of RSVP, they do. MPLS, per labeled path. Diffserv, per flow aggregate rather than per flow.

>   IMHO, just using DSCP bits on per-hop won't achieve 
>   QoS "Globally". You have to combine TE also.
>   That "QoS using TE" has been neglected by us.
>   That is what I am trying to convince Fred.

In my opinion, you need TE and you need _access control_ to achieve any credible QoS, no matter what scheme you use.

>   If you consider IP is a superset, MPLS, QoS and TE are
>   some overlapping subsets (pictorially):
> 
>        IP Superset
>        +-----------------------------------+
>        |     __________    __________      |
>        |    /          \  /           \    |
>        |   /   MPLS     \/    QoS      \   | 
>        |  /             /\              \  | 
>        | |             |  |             |  | 
>        | |             |  |             |  |
>        | |             |  |             |  |
>        | |          ___|__|___          |  |
>        |  \        /    \/    \        /   |
>        |   \      /     /\     \      /    |
>        |    \____/_____/  \_____\____/     |
>        |        |                |         |    
>        |        |                |         |
>        |        |                |         |
>        |         \      TE      /          |
>        |          \            /           |
>        |           \__________/            |
>        |                                   |
>        +-----------------------------------+
> 
>    You can just do MPLS (for fast label look-up etc).
>    You can achieve QoS using MPLS (without TE)
>    You can achieve better QoS using TE (with or 
>    without MPLS) etc etc...

I agree with the words you say above, but not the picture. Seems to me that QoS is a hoped-for result if you implement TE, MPLS, perhaps RSVP, perhaps Diffserv, and so on. QoS is not a technique, as MPLS and TE are. It is the result of using one or possibly several techniques.

>    IMHO, the intersection of all the subsets (ie., 
>    (QoS using MPLS-(DiffServ-Aware-TE) will make a 
>    *real-good* impact on end-to-end performance in the 
>    global Internet.

"Real good" would have to include access control (otherwise your QoS will go to hell anytime too many flows happen at an edge node, which can easily happen without controls on access) and also constraint-based routing (otherwise, the network will be apt to block new QoS service requests just because one possible route out of many might be temporarily congested). If the network core is assumed to be so fast that none of this is a concern, then the hope is that contraint-based routing is not so essential. I think that's where many of these QoS solutions seem to go. It gets "too hard" to do a complete solution. And a complete solution starts looking oh so similar to ATM.

Bert

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Wed Jul 24 15:14:32 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA28595
	for <routing-discussion-archive@odin.ietf.org>; Wed, 24 Jul 2002 15:14:32 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id PAA17782
	for routing-discussion-archive@odin.ietf.org; Wed, 24 Jul 2002 15:15:35 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA17010;
	Wed, 24 Jul 2002 15:06:55 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id PAA16980
	for <routing-discussion@optimus.ietf.org>; Wed, 24 Jul 2002 15:06:45 -0400 (EDT)
Received: from stl-smtpout-01.boeing.com (stl-smtpout-01.boeing.com [12.13.247.21])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA28248
	for <routing-discussion@ietf.org>; Wed, 24 Jul 2002 15:05:35 -0400 (EDT)
Received: from stl-av-02.boeing.com ([192.76.190.7])
	by stl-smtpout-01.boeing.com (8.9.2/8.8.5-M2) with ESMTP id OAA00194;
	Wed, 24 Jul 2002 14:06:35 -0500 (CDT)
Received: from blv-hub-01.boeing.com (localhost [127.0.0.1])
	by stl-av-02.boeing.com (8.9.3/8.9.2/MBS-AV-01) with ESMTP id OAA19364;
	Wed, 24 Jul 2002 14:06:23 -0500 (CDT)
Received: from xch-phl-02p.ne.nos.boeing.com (xch-phl-02p.ne.nos.boeing.com [128.225.22.4])
	by blv-hub-01.boeing.com (8.11.3/8.11.3/MBS-LDAP-01) with ESMTP id g6OJ6X629475;
	Wed, 24 Jul 2002 12:06:33 -0700 (PDT)
Received: by xch-phl-02p.ne.nos.boeing.com with Internet Mail Service (5.5.2650.21)
	id <37GWZQFP>; Wed, 24 Jul 2002 15:06:31 -0400
Message-ID: <8D6509B44ACCEB43A96E032B9C3EA0C8078A3A43@xch-ne-02.ne.nos.boeing.com>
From: "Manfredi, Albert E" <albert.e.manfredi@boeing.com>
To: "'Naidu, Venkata'" <Venkata.Naidu@marconi.com>,
        routing-discussion@ietf.org
Subject: RE: AW: Differentiated Routing, not only plain rambo-SPF
Date: Wed, 24 Jul 2002 15:06:25 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

> -----Original Message-----
> From: Naidu, Venkata [mailto:Venkata.Naidu@marconi.com]

> -> From: J. Noel Chiappa [mailto:jnc@ginger.lcs.mit.edu]

> -> 
> -> You just hit one of my hot buttons.
> -> 
> -> Normal IP routing is not stateless. It has lots of state in 
> -> the network.
>   
>   Hummel is not talking about *state* in the network.
>   The information (in your words: a Header) required to 
>   route packets is state-less. Routing is purely based on
>   *destination* address. Isn't it?

But the routers must maintain state information, such as network topology (OSPF) or spanning tree (RIP). True, under default IP best-effort routing, routers do not need to maintain state information *per flow*. But in case of RSVP, they do. MPLS, per labeled path. Diffserv, per flow aggregate rather than per flow.

>   IMHO, just using DSCP bits on per-hop won't achieve 
>   QoS "Globally". You have to combine TE also.
>   That "QoS using TE" has been neglected by us.
>   That is what I am trying to convince Fred.

In my opinion, you need TE and you need _access control_ to achieve any credible QoS, no matter what scheme you use.

>   If you consider IP is a superset, MPLS, QoS and TE are
>   some overlapping subsets (pictorially):
> 
>        IP Superset
>        +-----------------------------------+
>        |     __________    __________      |
>        |    /          \  /           \    |
>        |   /   MPLS     \/    QoS      \   | 
>        |  /             /\              \  | 
>        | |             |  |             |  | 
>        | |             |  |             |  |
>        | |             |  |             |  |
>        | |          ___|__|___          |  |
>        |  \        /    \/    \        /   |
>        |   \      /     /\     \      /    |
>        |    \____/_____/  \_____\____/     |
>        |        |                |         |    
>        |        |                |         |
>        |        |                |         |
>        |         \      TE      /          |
>        |          \            /           |
>        |           \__________/            |
>        |                                   |
>        +-----------------------------------+
> 
>    You can just do MPLS (for fast label look-up etc).
>    You can achieve QoS using MPLS (without TE)
>    You can achieve better QoS using TE (with or 
>    without MPLS) etc etc...

I agree with the words you say above, but not the picture. Seems to me that QoS is a hoped-for result if you implement TE, MPLS, perhaps RSVP, perhaps Diffserv, and so on. QoS is not a technique, as MPLS and TE are. It is the result of using one or possibly several techniques.

>    IMHO, the intersection of all the subsets (ie., 
>    (QoS using MPLS-(DiffServ-Aware-TE) will make a 
>    *real-good* impact on end-to-end performance in the 
>    global Internet.

"Real good" would have to include access control (otherwise your QoS will go to hell anytime too many flows happen at an edge node, which can easily happen without controls on access) and also constraint-based routing (otherwise, the network will be apt to block new QoS service requests just because one possible route out of many might be temporarily congested). If the network core is assumed to be so fast that none of this is a concern, then the hope is that contraint-based routing is not so essential. I think that's where many of these QoS solutions seem to go. It gets "too hard" to do a complete solution. And a complete solution starts looking oh so similar to ATM.

Bert

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Fri Jul 26 05:52:07 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA12814
	for <routing-discussion-archive@odin.ietf.org>; Fri, 26 Jul 2002 05:52:07 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA10520;
	Fri, 26 Jul 2002 05:47:34 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA10483
	for <routing-discussion@optimus.ietf.org>; Fri, 26 Jul 2002 05:47:32 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA12683
	for <routing-discussion@ietf.org>; Fri, 26 Jul 2002 05:46:26 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id LAA26692;
	Fri, 26 Jul 2002 11:47:06 +0200 (MET DST)
Received: from mchh273e.demchh201e.icn.siemens.de ([139.21.200.83])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id LAA29498;
	Fri, 26 Jul 2002 11:47:24 +0200 (MET DST)
Received: by MCHH273E with Internet Mail Service (5.5.2653.19)
	id <PP25Y5YS>; Fri, 26 Jul 2002 11:47:19 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B947@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'jshen@cad.zju.edu.cn'" <jshen@cad.zju.edu.cn>,
        "Naidu, Venkata"
	 <Venkata.Naidu@Marconi.com>
Cc: "'J. Noel Chiappa'" <jnc@ginger.lcs.mit.edu>, routing-discussion@ietf.org,
        "'fred@cisco.com'" <fred@cisco.com>
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Fri, 26 Jul 2002 11:47:19 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="gb2312"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Jing,
Yes, Differentiated Routing should be beneficial to both: a) service differentiation     A N D    b)  traffic balancing
(and later on, by learning from it, many other fancy services).

With ECMP however I have a lot of problems:
a) I don't like "equal-cost". Why is a little bit more costly route not suitable for becoming the 2nd or 3rd best route?
   This equal-cost goal only looks like a valuable goal, but in reality it just hides that the entire internet society is the 
   prisoner of this plain rambo-SPF routing paradigma.
b) It does not accomplish the wanted traffic load proportioning. See my email from last Wednesday. 
   No  policeman, who should balance the arriving car traffic, whould signal them which way to go based on 
   the U.S.-federal state on their car license plates.
c) Some flow/car which deserves best treatment may fall into the bad-luck hash category. So ECMP destroys DiffServ 
   intentions.

Jing wrote:
1. What's the target?
    I think the answer may be :  Differentiated service between flows + Load
balancing over networks + routing stability

HH=> Yes + Yes + Yes.
Yes to DiffServ effects: 
Better than a junction with traffic light (so that motor-cyclist can jump the queue of waiting cars)
is a ramp-based junction where the other traffic is kept out of the way. Remember I am proposing road systems
whereby you can pretty good ban some traffic onto some road system as to keep it off from others.
Yes to load balancing: 
Remember the road systems I proposed which are determined while taking into consideration the existence of previously determined road systems.
Yes to routing stability:
All the  packets of some flow come with some DSCP x and will therefore follow the same path (even if this DSCP x were
modified at some router to x-overflow for being forwarded on the x-overflow road system).  
  
 
Jing wrote:

2.  The basic method ?
   It may be :  multipath routing + DSCP based path selection  ( We do not
consider too much of CAC or the like here because we want to walk around
congestied  part of network by routing with multipath path )

HH=> to determine distinct road systems,  indexed by DSCP , implemented by the affected routers.

Jing wrote:
3. How could the overlaid network be built?
    I think this may be most difficult, as :  connection-oriented or
connectionless,  how could the path be computed, how could the path be
establised over network?
    MPLS may be a useful tool, but what about the scalability?

HH=> by deriving DSCP-x associated Routing Tables inside the affected routers.

Jing wrote:
4. Is there any confliction between  routing and  congestion control mechanism
in TCP ?
HH=> I cannot think of any at all.

Heinrich


 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Fri Jul 26 05:52:08 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA12822
	for <routing-discussion-archive@odin.ietf.org>; Fri, 26 Jul 2002 05:52:07 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id FAA10729
	for routing-discussion-archive@odin.ietf.org; Fri, 26 Jul 2002 05:53:12 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA10520;
	Fri, 26 Jul 2002 05:47:34 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA10483
	for <routing-discussion@optimus.ietf.org>; Fri, 26 Jul 2002 05:47:32 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA12683
	for <routing-discussion@ietf.org>; Fri, 26 Jul 2002 05:46:26 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id LAA26692;
	Fri, 26 Jul 2002 11:47:06 +0200 (MET DST)
Received: from mchh273e.demchh201e.icn.siemens.de ([139.21.200.83])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id LAA29498;
	Fri, 26 Jul 2002 11:47:24 +0200 (MET DST)
Received: by MCHH273E with Internet Mail Service (5.5.2653.19)
	id <PP25Y5YS>; Fri, 26 Jul 2002 11:47:19 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B947@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: "'jshen@cad.zju.edu.cn'" <jshen@cad.zju.edu.cn>,
        "Naidu, Venkata"
	 <Venkata.Naidu@Marconi.com>
Cc: "'J. Noel Chiappa'" <jnc@ginger.lcs.mit.edu>, routing-discussion@ietf.org,
        "'fred@cisco.com'" <fred@cisco.com>
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Fri, 26 Jul 2002 11:47:19 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2653.19)
Content-Type: text/plain;
	charset="gb2312"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Jing,
Yes, Differentiated Routing should be beneficial to both: a) service differentiation     A N D    b)  traffic balancing
(and later on, by learning from it, many other fancy services).

With ECMP however I have a lot of problems:
a) I don't like "equal-cost". Why is a little bit more costly route not suitable for becoming the 2nd or 3rd best route?
   This equal-cost goal only looks like a valuable goal, but in reality it just hides that the entire internet society is the 
   prisoner of this plain rambo-SPF routing paradigma.
b) It does not accomplish the wanted traffic load proportioning. See my email from last Wednesday. 
   No  policeman, who should balance the arriving car traffic, whould signal them which way to go based on 
   the U.S.-federal state on their car license plates.
c) Some flow/car which deserves best treatment may fall into the bad-luck hash category. So ECMP destroys DiffServ 
   intentions.

Jing wrote:
1. What's the target?
    I think the answer may be :  Differentiated service between flows + Load
balancing over networks + routing stability

HH=> Yes + Yes + Yes.
Yes to DiffServ effects: 
Better than a junction with traffic light (so that motor-cyclist can jump the queue of waiting cars)
is a ramp-based junction where the other traffic is kept out of the way. Remember I am proposing road systems
whereby you can pretty good ban some traffic onto some road system as to keep it off from others.
Yes to load balancing: 
Remember the road systems I proposed which are determined while taking into consideration the existence of previously determined road systems.
Yes to routing stability:
All the  packets of some flow come with some DSCP x and will therefore follow the same path (even if this DSCP x were
modified at some router to x-overflow for being forwarded on the x-overflow road system).  
  
 
Jing wrote:

2.  The basic method ?
   It may be :  multipath routing + DSCP based path selection  ( We do not
consider too much of CAC or the like here because we want to walk around
congestied  part of network by routing with multipath path )

HH=> to determine distinct road systems,  indexed by DSCP , implemented by the affected routers.

Jing wrote:
3. How could the overlaid network be built?
    I think this may be most difficult, as :  connection-oriented or
connectionless,  how could the path be computed, how could the path be
establised over network?
    MPLS may be a useful tool, but what about the scalability?

HH=> by deriving DSCP-x associated Routing Tables inside the affected routers.

Jing wrote:
4. Is there any confliction between  routing and  congestion control mechanism
in TCP ?
HH=> I cannot think of any at all.

Heinrich


 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Fri Jul 26 12:28:36 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28380
	for <routing-discussion-archive@odin.ietf.org>; Fri, 26 Jul 2002 12:28:35 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA05133;
	Fri, 26 Jul 2002 12:18:39 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA05064
	for <routing-discussion@optimus.ietf.org>; Fri, 26 Jul 2002 12:18:35 -0400 (EDT)
Received: from ernst.netinsight.se (mail1-2.netinsight.se [212.209.170.195])
	by ietf.org (8.9.1a/8.9.1a) with SMTP id MAA27865
	for <routing-discussion@ietf.org>; Fri, 26 Jul 2002 12:17:25 -0400 (EDT)
Received: from localhost (unverified [10.100.1.11]) by ernst.netinsight.se
 (EMWAC SMTPRS 0.83) with SMTP id <B0002009780@ernst.netinsight.se>;
 Fri, 26 Jul 2002 18:17:50 +0200
Date: Fri, 26 Jul 2002 18:17:49 +0200 (CEST)
Message-Id: <20020726.181749.74735582.magda@netinsight.net>
To: albert.e.manfredi@boeing.com
Cc: Venkata.Naidu@marconi.com, routing-discussion@ietf.org
Subject: Re: AW: Differentiated Routing, not only plain rambo-SPF
From: Magnus Danielson <magda@netinsight.net>
In-Reply-To: <8D6509B44ACCEB43A96E032B9C3EA0C8078A3A43@xch-ne-02.ne.nos.boeing.com>
References: <8D6509B44ACCEB43A96E032B9C3EA0C8078A3A43@xch-ne-02.ne.nos.boeing.com>
X-Mailer: Mew version 2.2 on Emacs 21.2 / Mule 5.0 (SAKAKI)
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Hi,

> > -> You just hit one of my hot buttons.
> > -> 
> > -> Normal IP routing is not stateless. It has lots of state in 
> > -> the network.
> >   
> >   Hummel is not talking about *state* in the network.
> >   The information (in your words: a Header) required to 
> >   route packets is state-less. Routing is purely based on
> >   *destination* address. Isn't it?

Nothing is purely stateless. The whole issue of the statelessness of IP (the
protocol or the suite of protocols is nowdays important to distinguish between)
is IMHO a historic relic. The real issue is how much state there is, where it
is and how we create and maintain it. The old scare of seeing state per flow is
still valid, since the number of flows on a large internet like the IP Internet
is very large, and this translates into a hell of a lot of memory and
processing. Since then it has been a debate on how best to cut and slice the
turkey into feasable chunks, even if it has not been spelled out.

> But the routers must maintain state information, such as network topology (OSPF) or spanning tree (RIP). True, under default IP best-effort routing, routers do not need to maintain state information *per flow*. But in case of RSVP, they do. MPLS, per labeled path. Diffserv, per flow aggregate rather than per flow.

Right. Question is, of all the methods, what actually works all the way?

> >   IMHO, just using DSCP bits on per-hop won't achieve 
> >   QoS "Globally". You have to combine TE also.
> >   That "QoS using TE" has been neglected by us.
> >   That is what I am trying to convince Fred.
> 
> In my opinion, you need TE and you need _access control_ to achieve any credible QoS, no matter what scheme you use.

Rather, if you are to guarantee a certain QoS, you need both to identify the
QoS needed, and means to within the network realize this. If you only have the
QoS requirement "fire and forget" then surely you do not need access control,
but the trouble comes when you want all of your transmitted bits to appear in
both good time and right order at the receiver(s), then you are really asking
for an allocation, and with allocations comes the need to limit these in order
to still ensure the service, and with that comes the need for policing and
this in turn put requirements on the transmitter and receiver sides in the
interaction with the network. Access control (actually, Call Access Control,
which is the right term in this scenario) is where this lead us. With such
strict interpretation of guarantees, we have allocation and thus "hard state"
(and not the so belowed "soft state").

In reality, I think that you can never avoid the hard guarantee and thus hard
allocation, nothing actually replaces it. However, hard guarantees is not
everything in a packet based network like the Internet.

The benefit of hard allocations is that the network now knows what to expect
from that part of the traffic. Just forget you had that capacity available.
This simplifies the Traffic Management and Traffic Engineering.

This is somewhat of a side-track, but I really think you can't acheive a full
system solution unless you first separate out the need for end-to-end strict
allocations and then deal with the rest. People tend to argue from very
different standpoints unless you do this.

> >    You can just do MPLS (for fast label look-up etc).
> >    You can achieve QoS using MPLS (without TE)
> >    You can achieve better QoS using TE (with or 
> >    without MPLS) etc etc...
> 
> I agree with the words you say above, but not the picture. Seems to me that QoS is a hoped-for result if you implement TE, MPLS, perhaps RSVP, perhaps Diffserv, and so on. QoS is not a technique, as MPLS and TE are. It is the result of using one or possibly several techniques.

Actually, QoS is defined to be a set of measures of the acheived Quality of the
service. Whatever techniques we put into use, they are there to somehow help
improve those measures, at least for the traffic which have a higher
requirement than other traffic.

If no diffrentiation in requirement between two sets traffic of traffic is
found, then they compete on the same level.

This whole soup is naturally complex, bitter-sweet and potentially leathal.

> >    IMHO, the intersection of all the subsets (ie., 
> >    (QoS using MPLS-(DiffServ-Aware-TE) will make a 
> >    *real-good* impact on end-to-end performance in the 
> >    global Internet.
> 
> "Real good" would have to include access control (otherwise your QoS will go
> to hell anytime too many flows happen at an edge node, which can easily
> happen without controls on access) and also constraint-based routing
> (otherwise, the network will be apt to block new QoS service requests just
> because one possible route out of many might be temporarily congested).

Again, blocking or not depends on which QoS goal you are attempting to acheive.
If you want strict guarantees you want yo be your own flow and you really want
your own capacity. This is when you want blocking. When you reach congestion
you have several methods to cope with it, like premtion by priority or by
having pushback mechanisms allowing thus that allows to be pushbacked to a
lower capacity do this in a controlled maner. When such measures fails,
blocking is what you want to get, since you could not aquire the guarantee you
requested.

You can then invent other softer allocation schemes, where blocking is not an
option, but where you never the less can get different treatments. Just
separating the treatment of normal TCP and that of UDP/RTP springs to mind.

> If the network core is assumed to be so fast that none of this is a concern,
> then the hope is that contraint-based routing is not so essential. I think
> that's where many of these QoS solutions seem to go. It gets "too hard" to
> do a complete solution. And a complete solution starts looking oh so similar
> to ATM.

Right, but on the same time one has to realize that it is not all about tossing
packets on this or this queue etc. The issue of time is really important too.
In ATM they had to synchronise the switches. Real-time traffic have (almost) by
definition tough requirements on the timing QoS measures, and this have
implications to how we route and handle the traffic, especially in relation to
other ("non-realtime traffic"). In reality we have a form of
pseudo/soft-realtime traffic (i.e. Web, FTP, Telnet/SSH, DNS) in addition to
hard realtime traffic (i.e. Audio/Video).

Also, for part of the traffic, there are requirments are allways end-to-end and
if a solution for that traffic is to exist on the Internet, then the solution
must work end-to-end, even if this will hurt in many places.

Sorry for this sidetrack, but IMHO it is often forgotten. Please continue with
this interesting thread.

Cheers,
Magnus

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Fri Jul 26 12:28:36 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28388
	for <routing-discussion-archive@odin.ietf.org>; Fri, 26 Jul 2002 12:28:36 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id MAA05804
	for routing-discussion-archive@odin.ietf.org; Fri, 26 Jul 2002 12:29:41 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA05133;
	Fri, 26 Jul 2002 12:18:39 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA05064
	for <routing-discussion@optimus.ietf.org>; Fri, 26 Jul 2002 12:18:35 -0400 (EDT)
Received: from ernst.netinsight.se (mail1-2.netinsight.se [212.209.170.195])
	by ietf.org (8.9.1a/8.9.1a) with SMTP id MAA27865
	for <routing-discussion@ietf.org>; Fri, 26 Jul 2002 12:17:25 -0400 (EDT)
Received: from localhost (unverified [10.100.1.11]) by ernst.netinsight.se
 (EMWAC SMTPRS 0.83) with SMTP id <B0002009780@ernst.netinsight.se>;
 Fri, 26 Jul 2002 18:17:50 +0200
Date: Fri, 26 Jul 2002 18:17:49 +0200 (CEST)
Message-Id: <20020726.181749.74735582.magda@netinsight.net>
To: albert.e.manfredi@boeing.com
Cc: Venkata.Naidu@marconi.com, routing-discussion@ietf.org
Subject: Re: AW: Differentiated Routing, not only plain rambo-SPF
From: Magnus Danielson <magda@netinsight.net>
In-Reply-To: <8D6509B44ACCEB43A96E032B9C3EA0C8078A3A43@xch-ne-02.ne.nos.boeing.com>
References: <8D6509B44ACCEB43A96E032B9C3EA0C8078A3A43@xch-ne-02.ne.nos.boeing.com>
X-Mailer: Mew version 2.2 on Emacs 21.2 / Mule 5.0 (SAKAKI)
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

Hi,

> > -> You just hit one of my hot buttons.
> > -> 
> > -> Normal IP routing is not stateless. It has lots of state in 
> > -> the network.
> >   
> >   Hummel is not talking about *state* in the network.
> >   The information (in your words: a Header) required to 
> >   route packets is state-less. Routing is purely based on
> >   *destination* address. Isn't it?

Nothing is purely stateless. The whole issue of the statelessness of IP (the
protocol or the suite of protocols is nowdays important to distinguish between)
is IMHO a historic relic. The real issue is how much state there is, where it
is and how we create and maintain it. The old scare of seeing state per flow is
still valid, since the number of flows on a large internet like the IP Internet
is very large, and this translates into a hell of a lot of memory and
processing. Since then it has been a debate on how best to cut and slice the
turkey into feasable chunks, even if it has not been spelled out.

> But the routers must maintain state information, such as network topology (OSPF) or spanning tree (RIP). True, under default IP best-effort routing, routers do not need to maintain state information *per flow*. But in case of RSVP, they do. MPLS, per labeled path. Diffserv, per flow aggregate rather than per flow.

Right. Question is, of all the methods, what actually works all the way?

> >   IMHO, just using DSCP bits on per-hop won't achieve 
> >   QoS "Globally". You have to combine TE also.
> >   That "QoS using TE" has been neglected by us.
> >   That is what I am trying to convince Fred.
> 
> In my opinion, you need TE and you need _access control_ to achieve any credible QoS, no matter what scheme you use.

Rather, if you are to guarantee a certain QoS, you need both to identify the
QoS needed, and means to within the network realize this. If you only have the
QoS requirement "fire and forget" then surely you do not need access control,
but the trouble comes when you want all of your transmitted bits to appear in
both good time and right order at the receiver(s), then you are really asking
for an allocation, and with allocations comes the need to limit these in order
to still ensure the service, and with that comes the need for policing and
this in turn put requirements on the transmitter and receiver sides in the
interaction with the network. Access control (actually, Call Access Control,
which is the right term in this scenario) is where this lead us. With such
strict interpretation of guarantees, we have allocation and thus "hard state"
(and not the so belowed "soft state").

In reality, I think that you can never avoid the hard guarantee and thus hard
allocation, nothing actually replaces it. However, hard guarantees is not
everything in a packet based network like the Internet.

The benefit of hard allocations is that the network now knows what to expect
from that part of the traffic. Just forget you had that capacity available.
This simplifies the Traffic Management and Traffic Engineering.

This is somewhat of a side-track, but I really think you can't acheive a full
system solution unless you first separate out the need for end-to-end strict
allocations and then deal with the rest. People tend to argue from very
different standpoints unless you do this.

> >    You can just do MPLS (for fast label look-up etc).
> >    You can achieve QoS using MPLS (without TE)
> >    You can achieve better QoS using TE (with or 
> >    without MPLS) etc etc...
> 
> I agree with the words you say above, but not the picture. Seems to me that QoS is a hoped-for result if you implement TE, MPLS, perhaps RSVP, perhaps Diffserv, and so on. QoS is not a technique, as MPLS and TE are. It is the result of using one or possibly several techniques.

Actually, QoS is defined to be a set of measures of the acheived Quality of the
service. Whatever techniques we put into use, they are there to somehow help
improve those measures, at least for the traffic which have a higher
requirement than other traffic.

If no diffrentiation in requirement between two sets traffic of traffic is
found, then they compete on the same level.

This whole soup is naturally complex, bitter-sweet and potentially leathal.

> >    IMHO, the intersection of all the subsets (ie., 
> >    (QoS using MPLS-(DiffServ-Aware-TE) will make a 
> >    *real-good* impact on end-to-end performance in the 
> >    global Internet.
> 
> "Real good" would have to include access control (otherwise your QoS will go
> to hell anytime too many flows happen at an edge node, which can easily
> happen without controls on access) and also constraint-based routing
> (otherwise, the network will be apt to block new QoS service requests just
> because one possible route out of many might be temporarily congested).

Again, blocking or not depends on which QoS goal you are attempting to acheive.
If you want strict guarantees you want yo be your own flow and you really want
your own capacity. This is when you want blocking. When you reach congestion
you have several methods to cope with it, like premtion by priority or by
having pushback mechanisms allowing thus that allows to be pushbacked to a
lower capacity do this in a controlled maner. When such measures fails,
blocking is what you want to get, since you could not aquire the guarantee you
requested.

You can then invent other softer allocation schemes, where blocking is not an
option, but where you never the less can get different treatments. Just
separating the treatment of normal TCP and that of UDP/RTP springs to mind.

> If the network core is assumed to be so fast that none of this is a concern,
> then the hope is that contraint-based routing is not so essential. I think
> that's where many of these QoS solutions seem to go. It gets "too hard" to
> do a complete solution. And a complete solution starts looking oh so similar
> to ATM.

Right, but on the same time one has to realize that it is not all about tossing
packets on this or this queue etc. The issue of time is really important too.
In ATM they had to synchronise the switches. Real-time traffic have (almost) by
definition tough requirements on the timing QoS measures, and this have
implications to how we route and handle the traffic, especially in relation to
other ("non-realtime traffic"). In reality we have a form of
pseudo/soft-realtime traffic (i.e. Web, FTP, Telnet/SSH, DNS) in addition to
hard realtime traffic (i.e. Audio/Video).

Also, for part of the traffic, there are requirments are allways end-to-end and
if a solution for that traffic is to exist on the Internet, then the solution
must work end-to-end, even if this will hurt in many places.

Sorry for this sidetrack, but IMHO it is often forgotten. Please continue with
this interesting thread.

Cheers,
Magnus

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Fri Jul 26 17:26:32 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA09535
	for <routing-discussion-archive@odin.ietf.org>; Fri, 26 Jul 2002 17:26:32 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA23984;
	Fri, 26 Jul 2002 17:25:24 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA12334
	for <routing-discussion@ns.ietf.org>; Thu, 25 Jul 2002 22:28:03 -0400 (EDT)
Received: from mail.cad.zju.edu.cn (cad.zju.edu.cn [210.32.131.2])
	by ietf.org (8.9.1a/8.9.1a) with SMTP id WAA27464
	for <routing-discussion@ietf.org>; Thu, 25 Jul 2002 22:26:58 -0400 (EDT)
Received: (qmail 30346 invoked from network); 26 Jul 2002 02:32:47 -0000
Received: from unknown (HELO cad.zju.edu.cn) (210.32.131.97)
  by 210.32.131.2 with SMTP; 26 Jul 2002 02:32:47 -0000
Message-ID: <3D40B391.9B8829B2@cad.zju.edu.cn>
Date: Fri, 26 Jul 2002 10:27:29 +0800
From: shen jing <jshen@cad.zju.edu.cn>
Reply-To: jshen@cad.zju.edu.cn
Organization: State Key Lab of CAD&CG
X-Mailer: Mozilla 4.79 [en] (X11; U; SunOS 5.6 sun4u)
X-Accept-Language: en
MIME-Version: 1.0
To: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
CC: "'J. Noel Chiappa'" <jnc@ginger.lcs.mit.edu>, routing-discussion@ietf.org,
        "'fred@cisco.com'" <fred@cisco.com>, Heinrich.Hummel@icn.siemens.de
Subject: Re: AW: Differentiated Routing, not only plain rambo-SPF
References: <39469E08BD83D411A3D900204840EC55763278@vie-msgusr-01.dc.fore.com>
Content-Type: text/plain; charset=gb2312
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

IMO,  Hummel present a very interesting question: how could traffic be routed
according to their DiffServ classes?

ECMP or OMP has been investigated in past days,  and simulation results showed
they can improve network
performance under the experiment conditions. If I do not make a mistake, there
is someone publishing a paper
on "alternative routing with DSCP" in "Computer Network". In that paper, the
author stated to reroute
traffic according to DSCP in packet header when congestion happens along the
primary path.

On the other side, routing is the basic mechanism to carry traffic through the
network, so it is one the key factors for  load balancing (TE)
or QoS.Without cooperation between routing , CAC and schedulting ,  QoS is
impossible, while load balancing may be done with
routing.  Starting from this, I think it should not be the center of
consideration whether IP routing is stateless or the contrary. We must
make clear what we want to achieve first.

It seems Hummel thought about "multipath routing with DSCP network " overlaid
above IP networks.  I think there may be some
very diffcult problem with such a idea:

1. What's the target?
    I think the answer may be :  Differentiated service between flows + Load
balancing over networks + routing stability

2.  The basic method ?
   It may be :  multipath routing + DSCP based path selection  ( We do not
consider too much of CAC or the like here because we want to walk around
congestied
   part of network by routing with multipath path )

3. How could the overlaid network be built?
    I think this may be most difficult, as :  connection-oriented or
connectionless,  how could the path be computed, how could the path be
establised over network?
    MPLS may be a useful tool, but what about the scalability?

4. Is there any confliction between  routing and  congestion control mechanism
in TCP ?






> Noel,
>
> -> From: J. Noel Chiappa [mailto:jnc@ginger.lcs.mit.edu]
> ->     > From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
> ->
> ->     > In normal, state-less IP routing
> ->
> -> You just hit one of my hot buttons.
> ->
> -> Normal IP routing is not stateless. It has lots of state in
> -> the network.
>
>   Hummel is not talking about *state* in the network.
>   The information (in your words: a Header) required to
>   route packets is state-less. Routing is purely based on
>   *destination* address. Isn't it?

[snip]



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Fri Jul 26 17:26:32 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA09543
	for <routing-discussion-archive@odin.ietf.org>; Fri, 26 Jul 2002 17:26:32 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id RAA24129
	for routing-discussion-archive@odin.ietf.org; Fri, 26 Jul 2002 17:27:38 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id RAA23984;
	Fri, 26 Jul 2002 17:25:24 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id WAA12334
	for <routing-discussion@ns.ietf.org>; Thu, 25 Jul 2002 22:28:03 -0400 (EDT)
Received: from mail.cad.zju.edu.cn (cad.zju.edu.cn [210.32.131.2])
	by ietf.org (8.9.1a/8.9.1a) with SMTP id WAA27464
	for <routing-discussion@ietf.org>; Thu, 25 Jul 2002 22:26:58 -0400 (EDT)
Received: (qmail 30346 invoked from network); 26 Jul 2002 02:32:47 -0000
Received: from unknown (HELO cad.zju.edu.cn) (210.32.131.97)
  by 210.32.131.2 with SMTP; 26 Jul 2002 02:32:47 -0000
Message-ID: <3D40B391.9B8829B2@cad.zju.edu.cn>
Date: Fri, 26 Jul 2002 10:27:29 +0800
From: shen jing <jshen@cad.zju.edu.cn>
Reply-To: jshen@cad.zju.edu.cn
Organization: State Key Lab of CAD&CG
X-Mailer: Mozilla 4.79 [en] (X11; U; SunOS 5.6 sun4u)
X-Accept-Language: en
MIME-Version: 1.0
To: "Naidu, Venkata" <Venkata.Naidu@Marconi.com>
CC: "'J. Noel Chiappa'" <jnc@ginger.lcs.mit.edu>, routing-discussion@ietf.org,
        "'fred@cisco.com'" <fred@cisco.com>, Heinrich.Hummel@icn.siemens.de
Subject: Re: AW: Differentiated Routing, not only plain rambo-SPF
References: <39469E08BD83D411A3D900204840EC55763278@vie-msgusr-01.dc.fore.com>
Content-Type: text/plain; charset=gb2312
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org
Content-Transfer-Encoding: 7bit

IMO,  Hummel present a very interesting question: how could traffic be routed
according to their DiffServ classes?

ECMP or OMP has been investigated in past days,  and simulation results showed
they can improve network
performance under the experiment conditions. If I do not make a mistake, there
is someone publishing a paper
on "alternative routing with DSCP" in "Computer Network". In that paper, the
author stated to reroute
traffic according to DSCP in packet header when congestion happens along the
primary path.

On the other side, routing is the basic mechanism to carry traffic through the
network, so it is one the key factors for  load balancing (TE)
or QoS.Without cooperation between routing , CAC and schedulting ,  QoS is
impossible, while load balancing may be done with
routing.  Starting from this, I think it should not be the center of
consideration whether IP routing is stateless or the contrary. We must
make clear what we want to achieve first.

It seems Hummel thought about "multipath routing with DSCP network " overlaid
above IP networks.  I think there may be some
very diffcult problem with such a idea:

1. What's the target?
    I think the answer may be :  Differentiated service between flows + Load
balancing over networks + routing stability

2.  The basic method ?
   It may be :  multipath routing + DSCP based path selection  ( We do not
consider too much of CAC or the like here because we want to walk around
congestied
   part of network by routing with multipath path )

3. How could the overlaid network be built?
    I think this may be most difficult, as :  connection-oriented or
connectionless,  how could the path be computed, how could the path be
establised over network?
    MPLS may be a useful tool, but what about the scalability?

4. Is there any confliction between  routing and  congestion control mechanism
in TCP ?






> Noel,
>
> -> From: J. Noel Chiappa [mailto:jnc@ginger.lcs.mit.edu]
> ->     > From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
> ->
> ->     > In normal, state-less IP routing
> ->
> -> You just hit one of my hot buttons.
> ->
> -> Normal IP routing is not stateless. It has lots of state in
> -> the network.
>
>   Hummel is not talking about *state* in the network.
>   The information (in your words: a Header) required to
>   route packets is state-less. Routing is purely based on
>   *destination* address. Isn't it?

[snip]



_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Tue Jul 30 05:05:22 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA09330
	for <routing-discussion-archive@odin.ietf.org>; Tue, 30 Jul 2002 05:05:22 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA11956;
	Tue, 30 Jul 2002 05:00:22 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA11923
	for <routing-discussion@optimus.ietf.org>; Tue, 30 Jul 2002 05:00:20 -0400 (EDT)
Received: from webmail22.rediffmail.com (webmail22.rediffmail.com [203.199.83.144] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with SMTP id EAA09128
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 04:59:05 -0400 (EDT)
Received: (qmail 27127 invoked by uid 510); 30 Jul 2002 09:00:19 -0000
Date: 30 Jul 2002 09:00:19 -0000
Message-ID: <20020730090019.27126.qmail@webmail22.rediffmail.com>
Received: from unknown (203.200.20.226) by rediffmail.com via HTTP; 30 Jul 2002 09:00:19 -0000
MIME-Version: 1.0
From: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
Reply-To: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
To: routing-discussion@ietf.org
Subject: Problem in RIP route summarization
Content-type: text/plain;
	format=flowed
Content-Disposition: inline
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hello all,

  I have problem in RIP route summarization.	As per RFC 2453, 
auto-summarization is done in order to reduce the total number of 
route entries. But Cisco also does summarization which is on 
interface basis. Will it be advisable to do route summarization on 
interfaces basis also, as this was not discussed in RFC?

I have some contradictions in Cisco implementation of 
summarization on interfaces basis. Some of the issues are as 
discussed below. All the references are picked up from Cisco 
document

http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/ip_c/ipcprt2/1cdrip.htm

Interface summary address overrides the Auto Summary address.
 	The AutoSummary will override the configured summary-address 
feature on a given interface except when the following condition 
is true:

-	The configured interface summary-address and the IP address of 
the configured interface share the same major network (the 
classful, nonsubnetted portion of the IP address).
-	Split horizon is not enabled on the interface.

Issue 1: On what basis the ip summary address overrides the auto 
summary address, if the major net of interface address and ip 
summary address matches?

Issue 2: On what scenario, there can be a problem when split 
horizon and auto-summary are enabled?

Issue 3: There is a contradiction is Cisco document saying 
auto-summary and split horizon (excluding non-broadcast networks 
)is enabled by default. But the primary reason for 
auto-summarization to come into effect is split horizon should be 
disabled.

I would appreciate if I can get solutions addressing above 
issues.

Thanks in advance.

With Regards,
S.Karthikeyan.

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Tue Jul 30 05:05:22 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA09339
	for <routing-discussion-archive@odin.ietf.org>; Tue, 30 Jul 2002 05:05:22 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id FAA12708
	for routing-discussion-archive@odin.ietf.org; Tue, 30 Jul 2002 05:06:30 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA11956;
	Tue, 30 Jul 2002 05:00:22 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id FAA11923
	for <routing-discussion@optimus.ietf.org>; Tue, 30 Jul 2002 05:00:20 -0400 (EDT)
Received: from webmail22.rediffmail.com (webmail22.rediffmail.com [203.199.83.144] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with SMTP id EAA09128
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 04:59:05 -0400 (EDT)
Received: (qmail 27127 invoked by uid 510); 30 Jul 2002 09:00:19 -0000
Date: 30 Jul 2002 09:00:19 -0000
Message-ID: <20020730090019.27126.qmail@webmail22.rediffmail.com>
Received: from unknown (203.200.20.226) by rediffmail.com via HTTP; 30 Jul 2002 09:00:19 -0000
MIME-Version: 1.0
From: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
Reply-To: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
To: routing-discussion@ietf.org
Subject: Problem in RIP route summarization
Content-type: text/plain;
	format=flowed
Content-Disposition: inline
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hello all,

  I have problem in RIP route summarization.	As per RFC 2453, 
auto-summarization is done in order to reduce the total number of 
route entries. But Cisco also does summarization which is on 
interface basis. Will it be advisable to do route summarization on 
interfaces basis also, as this was not discussed in RFC?

I have some contradictions in Cisco implementation of 
summarization on interfaces basis. Some of the issues are as 
discussed below. All the references are picked up from Cisco 
document

http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/ip_c/ipcprt2/1cdrip.htm

Interface summary address overrides the Auto Summary address.
 	The AutoSummary will override the configured summary-address 
feature on a given interface except when the following condition 
is true:

-	The configured interface summary-address and the IP address of 
the configured interface share the same major network (the 
classful, nonsubnetted portion of the IP address).
-	Split horizon is not enabled on the interface.

Issue 1: On what basis the ip summary address overrides the auto 
summary address, if the major net of interface address and ip 
summary address matches?

Issue 2: On what scenario, there can be a problem when split 
horizon and auto-summary are enabled?

Issue 3: There is a contradiction is Cisco document saying 
auto-summary and split horizon (excluding non-broadcast networks 
)is enabled by default. But the primary reason for 
auto-summarization to come into effect is split horizon should be 
disabled.

I would appreciate if I can get solutions addressing above 
issues.

Thanks in advance.

With Regards,
S.Karthikeyan.

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Tue Jul 30 12:55:05 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28331
	for <routing-discussion-archive@odin.ietf.org>; Tue, 30 Jul 2002 12:55:05 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA08512;
	Tue, 30 Jul 2002 12:43:40 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA08487
	for <routing-discussion@optimus.ietf.org>; Tue, 30 Jul 2002 12:43:38 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA27512
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 12:42:30 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id SAA14499
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 18:43:18 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id SAA14186
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 18:43:35 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2656.59)
	id <PTSY9608>; Tue, 30 Jul 2002 18:43:27 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B95B@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Tue, 30 Jul 2002 18:43:26 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2656.59)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Let's be concentrated again. And let me get to the heart of the matter.

Obviously I am not the only one who thought about using the DSCP also for doing Differentiated Routing.
Will say: not the only one who thought about indexing different Routing Tables by different DSCPs.

IMO, these routing tables should represent different and very distinct and differentiated road systems.
This is only possible, if they are genuine partial structures of the total network. In addition to them we still need
that "default"-road system which is identical to the entire network (e.g. in order to ping any node), i.e. today's routing tables.

As a first approach which already highlights all important aspects, let`'s consider just those "edge" nodes with attached
LANs, hosts,external interfaces. The task is to build a minimal road system between them, which suffices some well-determined
QoS/BW/Policy constraints, and which is able to lock the entire traffic between these edge nodes onto/into this road system
associated to a specific DSCP.

The graph of this minimal road system is certainly a "minimal size tree" (and not any shortest path tree with a very specific
root node). A couple of IETF meetings ago,I showed how such a tree can be constructed. At that time I was focussed on
determining a minimal partial mesh between n sites of a VPN. And the "big" gain: 10 % or 15 % less links than any shortest path tree between these n sites. In the meantime however, I realized a much bigger gain: It should be possible that 
each network node may compute the identical graph of some shortest size tree. All it takes is to apply some tie breaker rules
so that each node will compute the identical graph, even if there were several trees of smallest size.
Next:
As soon as the router has determined this graph, it checks, where is its own position relative to this graph.
If it is not touched, it won't have to derive a respective routing table. Otherwise however, it should derive one.
For deriving this routing table, the router runs the Dijkstra while taking into consideration only the links of this
smallest size tree.By doing so, it will get for each destination (prefix) what is its proper next hop link.

There will certainly be alternative road system graphs in addition to the minimal size tree graph. 
E.g. minimal rings, or even any kind of some meshy graph. Important is, that a particular road system turns out to be a "routable" road system, i.e. a road system where the edge nodes may reach each other without leaving this road system. 

A short and rough description on how to determine the shortest size tree between n edge nodes:
make n-1 iteration steps. At each step combine two partitions (of the tree). Start with n partitions of the tree whereby each
partion contains just a different edge node (not one single link). 
At each step combine those two partions which can be combined by the shortest path between any two nodes, 
whereby the one node is in this partition and the other node is in the other partition.
More iteration details later.
Before you start this iteration, you should have computed the shortest-path from any edge/core node to any edge/core node
while obeying the particular QoS/BW/Policy constraints and may store them in a kind of bin  from where you can take them
during the iteration.

I hope this explains why I look at SPF as a rambo routing protocol. Yes, SPF has its advantage. It get's you everywhere.
And also: for constructing the Smallest Size Tree, you even have to compute n Disjktra trees for filling the bin.
However, in the end, each router should determine the identical graph and look for its position rather than behave as the
center(=root) of the network.

Currently I can only see or feel how Differentiated Routing would work in an OSPF-network, and would need help and would appreciate any help to place this concept into a BGP environment, so that the road system may go somehow end-to-end.

Maybe the following is a nice comparison: there are airlines, which form alliances as to extend their flight road systems,
eventually such that you can fly end-to-end with one airline alliance. But note, you may as well travel with a different
airline alliance! Couldn't this be a business model also for ISPs ?

 
I hope, this get's the discussion back on its track. And away from the "what means state-less" debate - although I am grateful for any comment :-)

Heinrich





-----Ursprüngliche Nachricht-----
Von: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
Gesendet: Dienstag, 23. Juli 2002 15:19
An: 'Naidu, Venkata'; 'Fred Baker'; 'svenkata@gmu.edu'
Cc: routing-discussion@ietf.org
Betreff: AW: Differentiated Routing, not only plain rambo-SPF


Hi all debaters,
Thanks for this echo on my initial email yesterday evening, European time.
May I conclude that there is indeed a lot of interest in DiffRout  (if nsis exists as a WG, but
depends on Differentiated Routing (DiffRout), then we should have a diffrout WG, too:-)

Certainly, the basic idea of DiffRout is to use the DSCP-Bits, or even all 8 TOS-bits, for addressing different
Routing Tables for looking up the next hop link (you can as well have one common Routing Table but search for the right entry while using destination address AND DSCP as search keys - this is up to the programmer).
 
The big question however is, what shall the Routing Tables, inside of multiple routers and accessed by the same DSCP, represent?
My thinking is, they should implement a particular (logical) road system - next to other road systems on top of a given
physical network infrastructure.

Some of these road systems may have overlapping links. Some of them
must not have overlapping links. Some may serve different users or different services. Let me indulge in  analogies:

A road system for pedestrians comprises side-walks, sky-walks, streets, country roads, state roads, but not highways and express ways.
A road systems for cars comprises streets, country roads, state roads, highways, express ways, conditionally toll roads, but definitively no side walks and no sky walks (only in Hollywood movies).
A road system towards downtown might also have an associated overflow road system, which is a generalized form of a detour.
And even the overflow road system may have its overflow road system...
You see, some road system may even be defined based on/dependent on the existence of other road systems.

Traffic towards downtown using DSCP X may, at the rim of downtown, be redirected, by also changing DSCP X to DSCP X-overflow, if momentarily and locally available traffic load measurement data do advise so.  

We may consider a couple of  overflow road systems whereby, e.g. the forth one, is determined by disregarding the roads of the first three road systems. These four road systems may be used alternatively so that traffic is well balanced, i.e. so that overflow situations will not occur in the first place.And yet all packets of any given micro flow would use one of these four DSCPs, so that they would all take the same route.

Or consider a road system (=detour route)which is  just a single route tunnel from one road system to the other. And if a detour route between San Francisco and L.A. is associated with DSCP=5, no one would bother if another detour route road system between N.Y. and Baltimore were associated with the same DSCP=5.

We may even allocate bandwidth to some road system (and/or to its center )and may determine its links based on 
sufficiently available bandwidth (road system admission control !).

Also: If we LOCK e.g.the pedestrian traffic onto/into one road system we keep this traffic off the other road system
(and vice versa).

In summary: this is the way I look at it. And the goal should be  to develop solutions which implements such road systems:-)
What is your opinion?

Heinrich


 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Tue Jul 30 12:55:05 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28340
	for <routing-discussion-archive@odin.ietf.org>; Tue, 30 Jul 2002 12:55:05 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id MAA09544
	for routing-discussion-archive@odin.ietf.org; Tue, 30 Jul 2002 12:56:12 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA08512;
	Tue, 30 Jul 2002 12:43:40 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id MAA08487
	for <routing-discussion@optimus.ietf.org>; Tue, 30 Jul 2002 12:43:38 -0400 (EDT)
Received: from beamer.mchh.siemens.de (beamer.mchh.siemens.de [194.138.158.163])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA27512
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 12:42:30 -0400 (EDT)
Received: from moody.mchh.siemens.de (mail2.mchh.siemens.de [194.138.158.226])
	by beamer.mchh.siemens.de (8.9.3/8.9.3) with ESMTP id SAA14499
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 18:43:18 +0200 (MET DST)
Received: from mchh246e.demchh201e.icn.siemens.de ([139.21.200.56])
	by moody.mchh.siemens.de (8.9.1/8.9.1) with ESMTP id SAA14186
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 18:43:35 +0200 (MET DST)
Received: by MCHH246E with Internet Mail Service (5.5.2656.59)
	id <PTSY9608>; Tue, 30 Jul 2002 18:43:27 +0200
Message-ID: <EF8E39AA846CD411BB9C00508B951F510514B95B@MCHH267E>
From: Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>
To: routing-discussion@ietf.org
Subject: AW: Differentiated Routing, not only plain rambo-SPF
Date: Tue, 30 Jul 2002 18:43:26 +0200
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2656.59)
Content-Type: text/plain;
	charset="ISO-8859-1"
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Let's be concentrated again. And let me get to the heart of the matter.

Obviously I am not the only one who thought about using the DSCP also for doing Differentiated Routing.
Will say: not the only one who thought about indexing different Routing Tables by different DSCPs.

IMO, these routing tables should represent different and very distinct and differentiated road systems.
This is only possible, if they are genuine partial structures of the total network. In addition to them we still need
that "default"-road system which is identical to the entire network (e.g. in order to ping any node), i.e. today's routing tables.

As a first approach which already highlights all important aspects, let`'s consider just those "edge" nodes with attached
LANs, hosts,external interfaces. The task is to build a minimal road system between them, which suffices some well-determined
QoS/BW/Policy constraints, and which is able to lock the entire traffic between these edge nodes onto/into this road system
associated to a specific DSCP.

The graph of this minimal road system is certainly a "minimal size tree" (and not any shortest path tree with a very specific
root node). A couple of IETF meetings ago,I showed how such a tree can be constructed. At that time I was focussed on
determining a minimal partial mesh between n sites of a VPN. And the "big" gain: 10 % or 15 % less links than any shortest path tree between these n sites. In the meantime however, I realized a much bigger gain: It should be possible that 
each network node may compute the identical graph of some shortest size tree. All it takes is to apply some tie breaker rules
so that each node will compute the identical graph, even if there were several trees of smallest size.
Next:
As soon as the router has determined this graph, it checks, where is its own position relative to this graph.
If it is not touched, it won't have to derive a respective routing table. Otherwise however, it should derive one.
For deriving this routing table, the router runs the Dijkstra while taking into consideration only the links of this
smallest size tree.By doing so, it will get for each destination (prefix) what is its proper next hop link.

There will certainly be alternative road system graphs in addition to the minimal size tree graph. 
E.g. minimal rings, or even any kind of some meshy graph. Important is, that a particular road system turns out to be a "routable" road system, i.e. a road system where the edge nodes may reach each other without leaving this road system. 

A short and rough description on how to determine the shortest size tree between n edge nodes:
make n-1 iteration steps. At each step combine two partitions (of the tree). Start with n partitions of the tree whereby each
partion contains just a different edge node (not one single link). 
At each step combine those two partions which can be combined by the shortest path between any two nodes, 
whereby the one node is in this partition and the other node is in the other partition.
More iteration details later.
Before you start this iteration, you should have computed the shortest-path from any edge/core node to any edge/core node
while obeying the particular QoS/BW/Policy constraints and may store them in a kind of bin  from where you can take them
during the iteration.

I hope this explains why I look at SPF as a rambo routing protocol. Yes, SPF has its advantage. It get's you everywhere.
And also: for constructing the Smallest Size Tree, you even have to compute n Disjktra trees for filling the bin.
However, in the end, each router should determine the identical graph and look for its position rather than behave as the
center(=root) of the network.

Currently I can only see or feel how Differentiated Routing would work in an OSPF-network, and would need help and would appreciate any help to place this concept into a BGP environment, so that the road system may go somehow end-to-end.

Maybe the following is a nice comparison: there are airlines, which form alliances as to extend their flight road systems,
eventually such that you can fly end-to-end with one airline alliance. But note, you may as well travel with a different
airline alliance! Couldn't this be a business model also for ISPs ?

 
I hope, this get's the discussion back on its track. And away from the "what means state-less" debate - although I am grateful for any comment :-)

Heinrich





-----Ursprüngliche Nachricht-----
Von: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
Gesendet: Dienstag, 23. Juli 2002 15:19
An: 'Naidu, Venkata'; 'Fred Baker'; 'svenkata@gmu.edu'
Cc: routing-discussion@ietf.org
Betreff: AW: Differentiated Routing, not only plain rambo-SPF


Hi all debaters,
Thanks for this echo on my initial email yesterday evening, European time.
May I conclude that there is indeed a lot of interest in DiffRout  (if nsis exists as a WG, but
depends on Differentiated Routing (DiffRout), then we should have a diffrout WG, too:-)

Certainly, the basic idea of DiffRout is to use the DSCP-Bits, or even all 8 TOS-bits, for addressing different
Routing Tables for looking up the next hop link (you can as well have one common Routing Table but search for the right entry while using destination address AND DSCP as search keys - this is up to the programmer).
 
The big question however is, what shall the Routing Tables, inside of multiple routers and accessed by the same DSCP, represent?
My thinking is, they should implement a particular (logical) road system - next to other road systems on top of a given
physical network infrastructure.

Some of these road systems may have overlapping links. Some of them
must not have overlapping links. Some may serve different users or different services. Let me indulge in  analogies:

A road system for pedestrians comprises side-walks, sky-walks, streets, country roads, state roads, but not highways and express ways.
A road systems for cars comprises streets, country roads, state roads, highways, express ways, conditionally toll roads, but definitively no side walks and no sky walks (only in Hollywood movies).
A road system towards downtown might also have an associated overflow road system, which is a generalized form of a detour.
And even the overflow road system may have its overflow road system...
You see, some road system may even be defined based on/dependent on the existence of other road systems.

Traffic towards downtown using DSCP X may, at the rim of downtown, be redirected, by also changing DSCP X to DSCP X-overflow, if momentarily and locally available traffic load measurement data do advise so.  

We may consider a couple of  overflow road systems whereby, e.g. the forth one, is determined by disregarding the roads of the first three road systems. These four road systems may be used alternatively so that traffic is well balanced, i.e. so that overflow situations will not occur in the first place.And yet all packets of any given micro flow would use one of these four DSCPs, so that they would all take the same route.

Or consider a road system (=detour route)which is  just a single route tunnel from one road system to the other. And if a detour route between San Francisco and L.A. is associated with DSCP=5, no one would bother if another detour route road system between N.Y. and Baltimore were associated with the same DSCP=5.

We may even allocate bandwidth to some road system (and/or to its center )and may determine its links based on 
sufficiently available bandwidth (road system admission control !).

Also: If we LOCK e.g.the pedestrian traffic onto/into one road system we keep this traffic off the other road system
(and vice versa).

In summary: this is the way I look at it. And the goal should be  to develop solutions which implements such road systems:-)
What is your opinion?

Heinrich


 

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion

_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



From routing-discussion-admin@ietf.org  Wed Jul 31 18:55:21 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA21168
	for <routing-discussion-archive@odin.ietf.org>; Wed, 31 Jul 2002 18:55:20 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id SAA15467;
	Wed, 31 Jul 2002 18:52:22 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id EAA10629
	for <routing-discussion@optimus.ietf.org>; Tue, 30 Jul 2002 04:46:15 -0400 (EDT)
Received: from webmail22.rediffmail.com (webmail22.rediffmail.com [203.199.83.144] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with SMTP id EAA08747
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 04:45:01 -0400 (EDT)
Received: (qmail 12200 invoked by uid 510); 30 Jul 2002 08:39:15 -0000
Date: 30 Jul 2002 08:39:15 -0000
Message-ID: <20020730083915.12199.qmail@webmail22.rediffmail.com>
Received: from unknown (203.200.20.226) by rediffmail.com via HTTP; 30 Jul 2002 08:39:15 -0000
MIME-Version: 1.0
From: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
Reply-To: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
To: routing-discussion@ietf.org
Subject: Problem in RIP summarization
Content-type: text/plain;
	format=flowed
Content-Disposition: inline
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hello All,

    I have some problem in RIP route summarzation.As per RFC 2453, 
auto-summarization is done in order to reduce the total number of 
route entries. But Cisco also does summarization which is on 
interface basis. Will it be advisable to do route summarization on 
interfaces basis also, as this was not discussed in RFC?

I have some contradictions in Cisco implementation of 
summarization on interfaces basis. Some of the issues are as 
discussed below. All the references are picked up from Cisco 
document

http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/ip_c/ipcprt2/1cdrip.htm

Interface summary address overrides the Auto Summary address.
 	The AutoSummary will override the configured summary-address 
feature on a given interface except when the following condition 
is true:
-	The configured interface summary-address and the IP address of 
the configured interface share the same major network (the 
classful, nonsubnetted portion of the IP address).
-	Split horizon is not enabled on the interface.

Issue 1: On what basis the ip summary address overrides the auto 
summary address, if the major net of interface address and ip 
summary address matches?

Issue 2: On what scenario, there can be a problem when split 
horizon and auto-summary are enabled?

Issue 3: There is a contradiction is Cisco document saying 
auto-summary and split horizon (excluding non-broadcast networks 
)is enabled by default. But the primary reason for 
auto-summarization to come into effect is split horizon should be 
disabled.

I would appreciate if I can get solutions addressing above 
issues.

Thanks in advance.
With Regards,
S.Karthikeyan.


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion


From daemon@optimus.ietf.org  Wed Jul 31 18:55:23 2002
Received: from optimus.ietf.org (ietf.org [132.151.1.19] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA21182
	for <routing-discussion-archive@odin.ietf.org>; Wed, 31 Jul 2002 18:55:22 -0400 (EDT)
Received: (from daemon@localhost)
	by optimus.ietf.org (8.9.1a/8.9.1) id SAA15697
	for routing-discussion-archive@odin.ietf.org; Wed, 31 Jul 2002 18:56:28 -0400 (EDT)
Received: from optimus.ietf.org (localhost [127.0.0.1])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id SAA15467;
	Wed, 31 Jul 2002 18:52:22 -0400 (EDT)
Received: from ietf.org (odin [132.151.1.176])
	by optimus.ietf.org (8.9.1a/8.9.1) with ESMTP id EAA10629
	for <routing-discussion@optimus.ietf.org>; Tue, 30 Jul 2002 04:46:15 -0400 (EDT)
Received: from webmail22.rediffmail.com (webmail22.rediffmail.com [203.199.83.144] (may be forged))
	by ietf.org (8.9.1a/8.9.1a) with SMTP id EAA08747
	for <routing-discussion@ietf.org>; Tue, 30 Jul 2002 04:45:01 -0400 (EDT)
Received: (qmail 12200 invoked by uid 510); 30 Jul 2002 08:39:15 -0000
Date: 30 Jul 2002 08:39:15 -0000
Message-ID: <20020730083915.12199.qmail@webmail22.rediffmail.com>
Received: from unknown (203.200.20.226) by rediffmail.com via HTTP; 30 Jul 2002 08:39:15 -0000
MIME-Version: 1.0
From: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
Reply-To: "Karthikeyan Subramaniam" <keyans@rediffmail.com>
To: routing-discussion@ietf.org
Subject: Problem in RIP summarization
Content-type: text/plain;
	format=flowed
Content-Disposition: inline
Sender: routing-discussion-admin@ietf.org
Errors-To: routing-discussion-admin@ietf.org
X-Mailman-Version: 1.0
Precedence: bulk
List-Id: Routing Area General mailing list <routing-discussion.ietf.org>
X-BeenThere: routing-discussion@ietf.org

Hello All,

    I have some problem in RIP route summarzation.As per RFC 2453, 
auto-summarization is done in order to reduce the total number of 
route entries. But Cisco also does summarization which is on 
interface basis. Will it be advisable to do route summarization on 
interfaces basis also, as this was not discussed in RFC?

I have some contradictions in Cisco implementation of 
summarization on interfaces basis. Some of the issues are as 
discussed below. All the references are picked up from Cisco 
document

http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/ip_c/ipcprt2/1cdrip.htm

Interface summary address overrides the Auto Summary address.
 	The AutoSummary will override the configured summary-address 
feature on a given interface except when the following condition 
is true:
-	The configured interface summary-address and the IP address of 
the configured interface share the same major network (the 
classful, nonsubnetted portion of the IP address).
-	Split horizon is not enabled on the interface.

Issue 1: On what basis the ip summary address overrides the auto 
summary address, if the major net of interface address and ip 
summary address matches?

Issue 2: On what scenario, there can be a problem when split 
horizon and auto-summary are enabled?

Issue 3: There is a contradiction is Cisco document saying 
auto-summary and split horizon (excluding non-broadcast networks 
)is enabled by default. But the primary reason for 
auto-summarization to come into effect is split horizon should be 
disabled.

I would appreciate if I can get solutions addressing above 
issues.

Thanks in advance.
With Regards,
S.Karthikeyan.


_______________________________________________
routing-discussion mailing list
routing-discussion@ietf.org
https://www1.ietf.org/mailman/listinfo/routing-discussion



