RE: About the idea of layering iSCSI on VI on TCP

["Douglas Otis" <dotis@sanlight.net>]

Ralph,

Thanks for the information.  This T10 standard would represent a substantial
overhead for Ethernet as it was intended for Infiniband.  A VI approach
coupled with TCP is the only means iSCSI can exist.  It would be futile to
grope for headers and double handle memory without some other mechanism in
addition to a vanilla TCP, regardless of the desire.  The simple proposal
presented by Alex Nicolson from Giganet, albeit a bit late, looks far more
appropriate than the T10 spec.  The iSCSI proposal can not stand on its own.
iSCSI requires an extension to the present suite of IP protocols.  It would
not be appropriate to suggest security belongs to iSCSI because of the need
for this IP protocol extension.  It should belong to the protocol extension,
otherwise you are not creating a standard mechanism.  An extension to the
protocol that will be placed in hardware as well.  For David to discuss
getting rid of the design team to make progress underscores his
appreciation.

Doug

-----Original Message-----
From: owner-ips@ece.cmu.edu [mailto:owner-ips@ece.cmu.edu]On Behalf Of
Ralph Weber
Sent: Friday, July 28, 2000 10:30 AM
To: IPS Reflector
Subject: About the idea of layering iSCSI on VI on TCP


Forgive me if I've read the recent reflector traffic incorrectly,
but it appears that the idea is being floated of layering iSCSI
on VI on TCP.  Based on what I know about the activities of the
SCSI standards community as the secretary for T10 and the project
editor for SAM-2 and SPC-2 (SCSI Primary Commands) I'm wondering
if iSCSI would be needed in such a protocol stack.

T10 already has a project to layer SCSI on VI, it's call SVP
or SCSI VI Protocol.  The project editor for SVP is Ed Gardner
(Gardner@acm.org) and the project has a draft standard document
available at: ftp://ftp.t10.org/t10/drafts/svp/svp-r01.pdf

Recent SVP status report documents are available at:

  ftp://ftp.t10.org/t10/document.99/99-316r1.pdf
  ftp://ftp.t10.org/t10/document.00/00-172r0.pdf
  ftp://ftp.t10.org/t10/document.00/00-240r0.pdf
  ftp://ftp.t10.org/t10/document.00/00-278r0.pdf

This list is ordered by date from older to newer documents.

As far as I can tell, should the IPS WG decide to pursue
enabling SCSI over IP via definition of a VI protocol for
TCP, then that one document would be the only document
needed as the protocol stack would be SVP on VI on TCP.
Effectively, the proposed use of iSCSI would be replaced
by SVP.

Thanks.

Ralph...

P.S. I am aware of the IETF policy to use plain text files
and therefore apologize for the pointers to PDF files.  However,
the T10 policy is to use PDF files so there's no other choice.