RE: iSCSI Naming and Discovery

["John Hufferd/San Jose/IBM" <hufferd@us.ibm.com>]

Douglas Otis,
You said " ... You should consider the OS requirements for mounting.  The
database which defines the target:LUN:WWN should also include a name for
mounting...".

I know of no such "Requirement".  So lets get back to the real problems and
requirements.  I think the key point that we need to work on is the Naming
and Discovery of iSCSI Devices (Storage Controllers) in the Greater IP
network environment.

iSCSI Team,

I have been asked, why we are going over this Naming and Discovery stuff
now when we agreed to get the main protocol document out first.  That is
true, however, we have to assure ourselves that the Login process --perhaps
supplemented with the approprate "Text String" approach is sufficient to be
able to access the Storage Controllers in the IP network.  We also should
factor in the needs of "Edge or Boundary" Bridges/Gateways, and be sure
that they can also be appropriately supported by the Draft Proposal.  This
applies to both IP Bridges/Gateways as well as Bridges/Gateways from IP to
FC/SCSI.  We do not, at this time, need to nail down every thing associated
with Discovery and Naming to have sufficient information to insure the
Protocol Draft is adequate.

I have also been told that there is a major networking trend in which
customers are demanding methods of having "protocol private" networks.
These seem to be sub-networks with various forms of IP based Gateways,
which are protocol dependent, and  permit networks to be isolated to only
specific application protocols.  I must admit that I was surprised by this
trend, but I have heard something like it as folks have been concern about
"mixing" the iSCSI protocols with other protocols, on the same general
network.  Whether this is rational or not, is not as important as it is to
understand that it is an important trend and people want to have that
capability.

After detail talks with Costa, we reached a  understanding that the current
draft supported a general DNS approach for opening a TCP/IP connection
while resolving a Human understandable name through various normal Routers
and Switches to the ultimate target.  No changes seem to be needed to the
Draft Protocol to support that.

On the other hand,  in the notes that Costa (and Hafner) sent, which seemed
to be in contrast to the above, was actually focused on how to get through
the probable "Private-Network-Protocols-Specific-Gateways" which they
thought would be demanded, based on current trends.  They wanted to be sure
that there was enough information being exchanged at the establishment of
the iSCSI Session, that "Private-Network-Protocols-Specific-Gateways" could
in fact be built.  Hence, they were trying to define the approprate "TEXT"
String that such a device could use in that effort.  Costa, and I (and
perhaps Hafner and others) have now reach the conclusion that a separate
"Connect" command is probably not required.

In part of a conversation I had recently, I was told that the
Protocols-Specific-Gateways, are so simple to build that as soon as the
various vendors understand how to do it, they will roll out
Gateways/Routers that will perform that function within weeks to 3 months.
Also I was told that the overhead is so low in doing this, that generally
it was not a concern.  I guess that is because, the only real overhead is
on connection establishment, and not during normal flow, since tables are
set at connection establishment time which permits the normal execution
path to be very thin.  Now you can all believe that or not, but it sounds
right to me.

The finial consideration in this Gateway/Bridge focus will be analyzing the
needs of the iSCSI to FC and FC to iSCSI Routers/Switches (They are also
forms of Gateways/Bridges).  You may have seen some of that communication
with Joshua.

Again the key focus, in this thread (at this time), is to ensure that the
Draft Protocol has the right things in it that permit these various types
of routing to occur.





.
.
.
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


"Douglas Otis" <dotis@sanlight.net> on 10/06/2000 05:47:48 PM

To:   John Hufferd/San Jose/IBM@IBMUS, <ips@ece.cmu.edu>
cc:
Subject:  RE: iSCSI Naming and Discovery



John,

If I understand the point you are making...

You should consider the OS requirements for mounting.  The database which
defines the target:LUN:WWN should also include a name for mounting.
Perhaps
for Joe, it would be /loving_person.  So the LUN name for OS use, would
associate with the user so that once mounted, an application would be able
to find the data.  Most applications expect rather static mount locations.
It would be difficult to remember all of this should dozens of volumes are
mounted and unmounted on a rapid basis.  A network based system would allow
this to happen rather often so you should not overlook this requirement.
Perhaps only the port is verified should the access be directly to the
drive, but there should be a name:user definition maintained within the
access database.

Doug

> -----Original Message-----
> From: owner-ips@ece.cmu.edu [mailto:owner-ips@ece.cmu.edu]On Behalf Of
> John Hufferd/San Jose/IBM
> Sent: Friday, October 06, 2000 4:32 PM
> To: ips@ece.cmu.edu
> Subject: Re: iSCSI Naming and Discovery
>
>
> Jim Hafner,
> I wish you would NOT get "...down off the pulpit...".  This group seems
to
> need your preaching that LU names do NOT go into the Name Service.
Please
> keep it up.  I will look to you to be the anointed person to keep up the
> preaching, and the drum beat, so the rest of us can send our flames to
> other areas.  This is the area that you know cold, and I, and I hope the
> rest of the team,  will be looking to you for continuing guidance in this
> area as well as the related areas of  SCSI Access Controls, and Extended
> Copy etc.
>
> .
> .
> .
> 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
>
>
> Jim Hafner/Almaden/IBM@IBMUS@ece.cmu.edu on 10/06/2000 02:24:42 PM
>
> Sent by:  owner-ips@ece.cmu.edu
>
>
> To:   David Robinson <David.Robinson@EBay.Sun.COM>
> cc:   ips@ece.cmu.edu
> Subject:  Re: iSCSI Naming and Discovery
>
>
>
>
> David,
>
> You wrote:
>
>
> <snip>
> >The partial source of
> >my confusion is others wishing to lookup the LU in a nameservice,
> >not just the target.
>
> And that's the confusion I keep trying to clear up!  Any
> reference to LU in
> any nameservice is NOT needed! [I'll get down off the pulpit now, even
> though there's still a lot I can say here.]
>
> >Again, my fault, we are in complete agreement.
>
> Great!
> Jim Hafner
>
>
>
>
>