Re: iSCSI: SCTP, credits, proposals

[Randall Stewart <rrs@cisco.com>]

Black_David@emc.com wrote:

> Fortunately, we have an SCTP expert on this list
> (Randall Stewart) who should be able to flag things
> that are at odds with SCTP in a timely fashion.  The
> recent consensus to remove the command numbers
> and sliding windows from the standard header works
> well with SCTP, as use of multiple SCTP connections
> in a single iSCSI session seems rather unlikely.
>

David:

I will only acknowledge the title of expert in the sense
that:
a) Ex means former
   and
 b) spert is a drip..

In other words, I will own up to being a former drip :)


On the serious side of things... So far I see nothing in iSCSI that
does not make SCTP a minor drop in replacement for TCP. The
only thing is you would not want to put in multiple connections.
And since we will be getting a seperate document for multiple
connections, may I make a suggestion that this document
possibly also take up the use of multiple streams in place of
multiple connections when SCTP is used in place of TCP...

I will even volunteer to write some text for the document... my plate
is emptying out (don't tell anyone :->) and I will have some cycles I
can devote to this. I will be glad to help out in whatever  small ways
are possible for me to apply my limited talents :)


>
> The thread ending in the proposal of BB and EE credits
> was an unfortunate use of list bandwidth.  As was stated
> toward the end of that thread, these are flow control
> mechanisms that do not implement congestion control.
> I strongly suggest that everyone who has not already
> done so read RFC 2914, which is the RFC version of the
> floyd draft on congestion that I've referred to in the past.
> The next person who leads us down the rathole of asserting
> that flow control mechanisms solve congestion control problems
> should not expect a polite and accommodating response.
> Credits could form the basis of a congestion control
> mechanism, but one should expect to invest at least a
> year, probably longer in design and prototyping to get it
> to pass muster with the co-chairs, ADs, and IESG.
>

Yes, I think we have poured enough concrete in that Rat-Hole and
we have a consenses :)

R


> Thanks,
> --David
>
> ---------------------------------------------------
> David L. Black, Senior Technologist
> EMC Corporation, 42 South St., Hopkinton, MA  01748
> +1 (508) 435-1000 x75140     FAX: +1 (508) 497-8500
> black_david@emc.com       Mobile: +1 (978) 394-7754
> ---------------------------------------------------