RE: iSCSI: ISID issue

["Julian Satran" <Julian_Satran@il.ibm.com>]

David,

Only a slight correction to ease your conscience. The text calls for a
dynamic ISID allocated by the OS. Jim just suggested that
in the kickoff stage implementation can get away with an almost static
allocation and even later may use a dynamic partitioning scheme instead of
a recourse to OS on every session creation.

Julo


                                                                                               
                    Black_David@em                                                             
                    c.com                To:     John Hufferd/San Jose/IBM@IBMUS,              
                    Sent by:              ips@ece.cmu.edu                                      
                    owner-ips@ece.       cc:                                                   
                    cmu.edu              Subject:     RE: iSCSI: ISID issue                    
                                                                                               
                                                                                               
                    04-10-01 18:24                                                             
                    Please respond                                                             
                    to Black_David                                                             
                                                                                               
                                                                                               



John,

Lacking additional visible support, I'm going to give up on this
issue, but I think a serious mistake is being made.  Two portions
of your message reinforce my opinion:

> However, there were some other things that we batted around that I will
> cover here.  First, the Name which you want to work with in the iSCSI
ACLs
> is the iSCSI Initiator Node Name.  That should be the same for ALL iSCSI
> Initiators in a Host Node.  If vendors attempt to use anything else, I
> think they are not following the spec.  And should be sent to Singapore
for
> Caning :-)

Not :-), rather an unrealistic exercise in wishful thinking.  I have yet
to see a coherent explanation of why this will not repeat the Fibre Channel
Node WWN experience (which started out with the same intention and the same
"not following the spec" notion for use of different Node WWNs on the same
Platform - nonetheless, that's what happened).  I expect problems to arise
not so much in iSCSI ACLs (a new mechanism), but in the adaptation of
existing
SCSI LU access control mechanisms to iSCSI.

> When Jim talked about the new SCSI idea that would have Persistent
Reserves
> being tied to the Initiator Node,  it seemed clear to me, and I think to
> the others on the team, that the iSCSI Initiator Node Name was ideal to
be
> used for that purpose.  Jim pointed out, again to us, that this name was
> NOT part of any other Nexus meaning, it is only to be used with
Persistent
> Reserves.   All other Nexus meanings are against the iSCSI Initiator's
FULL
> NAME which is the  iSCSI initiator Node Name and the ISID.

That is a change from what Jim has described on the list, which required
a persistent name for the iSCSI Initiator Port (not Node); what you have
described in the above paragraph does not.  Given that this sort of change
is possible in what T10 is considering, I think we should cease trying to
anticipate and provide for what T10 might (or might not) do, and instead
deal with the results of what T10 does after it happens.

The upshot is that ISIDs now serve no visible purpose, aside from
complicating session setup - nonetheless, they're apparently going to
stay in the specification ... did the N&D folks really intend this result?

--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
---------------------------------------------------