Date: Sun, 27 Jul 2003 20:38:58 -0500
Subject: Re: iSCSI/iWARP drafts and flow control
Content-Type: text/plain; charset=US-ASCII; format=flowed
Mime-Version: 1.0 (Apple Message framework v552)
Cc: ips@ece.cmu.edu
To: Mike Ko <mako@almaden.ibm.com>
From: Caitlin Bestler <cait@asomi.com>
In-Reply-To: <OFE75635AC.5EBA9336-ON88256D71.000534FF@us.ibm.com>
Message-Id: <41FEB84F-C09C-11D7-902E-003065D48EE0@asomi.com>
Content-Transfer-Encoding: 7bit
Sender: owner-ips@ece.cmu.edu
Precedence: bulk


On Sunday, July 27, 2003, at 08:31 PM, Mike Ko wrote:

> Caitlin, it looks like as far as iSCSI asynchronous messages are
> concerned, we both agree that "setting limits on most of the types 
> listed
> is so easy that there is no need to negotiate the limit".  So the only
> other item is the SCSI asynchronous events, and the question yet to be
> answered is how many AENs can realistically be sent at one time.  Since
> the optional Shared Receive Queue already provides a solution to 
> handling
> this situation without impacting the iSCSI spec, does anyone else 
> thinks
> that the handling of AENs justify adding a negotiable item to set a 
> hard
> limit on the number of AENs?
>
> Mike Ko
> IBM Almaden Research
> San Jose, CA 95120

The DDP draft explicitly places responsibility for flow control of 
untagged
messages on the ULP.

That's "flow control", not "flow estimation".

"Control" ultimately involves *limiting* the amount that *may* be 
transmitted.
Just as the buffer limits do for TCP and SCTP. It is the fact that 
buffer
limits are effectively removed for RDMA streams that leads to the 
*requirement*
that the ULP perform flow control on untagged message.

So the question is whether the limit on the number of asynchronous 
messages
is specified in iSER or negotiated dynamically.

On *that* issue, I can't really think of any reason why there would ever
be a need to negotiate a non-standard amount.




Caitlin Bestler - cait@asomi.com - http://asomi.com/