RE: Re: iSCSI: Markers

["rakesh@qpackets.com" <rakesh@qpackets.com>]

My two cents worth in response to John's comments.

Before making any proposals and/or vote on choices we should be able to understand two things. 

1) why do we need framing?
2) based on the application (iscsi hardware or software assist) which framing technique will make more sense?

- Rakesh

John Hufferd wrote:

> OK, Folks, I have now talked to Steph, who authored  TUF, which is
> currently on the road to Experimental Status,  He has authored another
> version of TUF also, which uses a form of COWS.  So that means that we have
> two different versions of TUF as well as 2 versions of COWS (which are
> independent of Framing), and then there is FIM.  So let me list them and be
> sure we name them so that we are not in the middle of more confusion.
> 
> 1. Fixed Interval Markers (FIM) Currently In the iSCSI Draft
> 2. Constant Overhead Word Stuffing (COWS) as drafted by Julian and sent in
> his note of 12/23/2001 Subject "iSCSI - Synch an Steering Appendix -
> Markers & COWS"
> 3. TCP Upper-layer-protocol Framing (TUF) as drafted by Stephen Bailey in
> http://www.ietf.org/internet-drafts/draft-ietf-tsvwg-tcp-ulp-frame-01.txt
> 4. COWS Drafted By Stephen Bailey which can be used in both in stream and
> with Framing in
> http://www.cs.uchicago.edu/~steph/draft-bailey-tsvwg-cows-00.txt
> 
> Now lets call Julian's proposal COWS with 2 way pointers (COWS2WP)
> Now lets call Steph's COWS with 1 way pointers (COWS1WP)
> 
> When the type of COWS does not matter we can just call them COWS.
> 
> Both COWS can be used in Framing.  But to keep this discussion somewhat
> simpler lets call the Framing without any COWS as "Bare Framing", and Both
> of the other as "COWS Framing".  Only when we need to talk about which type
> of COWS should we say "COWS2WP Framing" or "COWS1WP Framing".  But for most
> conversation it should be just "COWS Framing".
> 
> So we have FIM, COWS1WP, COWS2WP, Bare Framing, & COWS Framing (made up of
> COWS1WP Framing and COWS2WP Framing).
> 
> Now we also need to understand that one of the main reasons expressed to
> make Framing go experimental, instead of Standards Track was that folks
> were worried that Bare Framing was based on probability, and that there was
> a very remote possibility that something could be done incorrectly.
> 
> As a result of that Steph was considering, as part of the experimental
> work, seeing what the impact of his previous COWS Draft would be on the
> experimental work that was going to be done.  He had no intention of
> bringing it up now, since he felt work/thought was still needed.
> 
> As you know COWS came up anyhow (and in a different form).
> 
> So what we have are statements from folks like me that had read Julian's
> Draft and the ietf-tsvwg version of Framing (Bare Framing), which did not
> see in those drafts the overlap.  Clearly there is an overlap in the minds
> of Julian for COWS2WP and Steph for COWS1WP and how they might impact
> Framing.
> 
> NET of Bare Framing vs COWS Framing:
> Bare Framing is based on probability and does not have to inspect each Word
> (SW or HW) COW requires Touching each Word,
> COWS Framing is guaranteed to always be correct.
> 
> So the choices are:
> 1. FIM now, and Bare Framing later
> 2. FIM now, and COWS Framing later
> 3. COWS now, and Bare Framing later
> 4. COWS now, and COWS Framing Later
> 5. Nothing now, and some kind of Framing Later
> 
> If we chose to do any of the "COWS now" options we would need to hold the
> debate on which form, but we should assume that which ever COWS we chose
> now is the COWS for later.
> 
> Value Statements
> 1. FIM and Bare Framing: Means we never have the overhead of touching every
> word
> 2. FIM and COWS Framing: Means that touching is postponed until Framing,
> and perhaps Faster Desktops/Laptops or support even support in normal NICs.
> 3.  COWS now and Bare Framing later: Has issues of toughing everything now,
> and then not useful later
> 4. COWS now and COWS Framing Later: Means always touch, but current
> approach is extensible into Framing
> 5.Nothing now, and some kind of Framing later: Means No current help, and
> no guarantee of help in the future, but some reasonable probability that
> some form of Framing will happen.
> 
> So it is 1-5 upon which  we should be taking a position.
> .
> .
> .
> John L. Hufferd
> Senior Technical Staff Member (STSM)
> IBM/SSG San Jose Ca
> Main Office (408) 256-0403, Tie: 276-0403,  eFax: (408) 904-4688
> Home Office (408) 997-6136, Cell: (408) 499-9702
> Internet address: hufferd@us.ibm.com
> 
> 

--------------------------------------------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .