RE: iSCSI virtualization

["sarel" <sarel@sanrad.com>]

I would have loved to agree with Yaron, that SCSI reflection status is an
iSCSI matter and should be discussed here and now, but in my opinion, this
feature should be “transport independent” thus being developed under T10. I
do agree that we might consider adding some hooks in the future to support
it as needed.

I would also like to mention that issuing additional I/O operation for each
I/O might be considered an unacceptable penalty in an I/O intensive
environment.

On the other hand, sharing thoughts and innovative ideas with the industry
should always be welcomed.


Sarel Altshuler
SANRAD

-----Original Message-----
From: John Hufferd/San Jose/IBM [mailto:hufferd@us.ibm.com]
Sent: Sunday, October 15, 2000 2:46 AM
To: ips@ece.cmu.edu
Subject: iSCSI virtualization


iSCSI Team,
With my TC Hat on:

The proposal, by Yarom, regarding virtualization has had some interesting
responses that have moved from support of the general idea to some
skepticism, to many folks wanting to delay the consideration to after the
main projects have been addressed.
Here are extracts of some of the messages (from earliest to latest and
excluding my own) :

I like the idea but am not so sure about the solution.(J.S)
I like  the basic idea of farming out the data transfer traffic to the real
devices. .... however, ...(C.M.)
I would suggest that this is an issue for T10.(J.H.)
How would discovery work in this scenario? (What would who "find"?)(D.D.)
This virtualization would make this more generally applicable,(M.B.)
The feature is very useful if it works transparently, within the framework
of existing device semantics and commands.(C.M.)
But at closer examination the case becomes weaker...(J.S.)
In my opinion,it's not [a T10 issue]. In any event, I vote for taking this
off the table.(C.M.)
I entirely agree that this is premature (Private)
I think this is an implementation issue (S.G.)
Let us keep it out of the iSCSI protocol for now. Maybe someday (S.G.)
Actually, the matter is T10 and T10 only. (R.S.)

I believe that the following is a fair statement of rough consensus:

1. The proposal, for virtualization, may have some merit.
2. We should take it off the iSCSI table (at least for now)

Therefore, based on the above, I am asking Yarom to create a Draft, which
can be fully debated, and submit it after the basic iSCSI protocol and
related projects are completed.

Hat off:

.
.
.
John L. Hufferd
Senior Technical Staff Member (STSM)
IBM/SSG San Jose Ca
(408) 256-0403, Tie: 276-0403
Internet address: hufferd@us.ibm.com