RE: iSCSI Naming and Discovery

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

Douglas Otis,
Again, I think you have invented this need.  Yes there maybe new
applications that may find a need to do what you said, but I am not sure
that we need to worry about them, since they will be very few in number and
can save what ever application information they need where ever they want
to save it.

The reason I say this, is because we are NOT building a NAS with a Shared
File System on the Target, we are talking about "Raw" Volumes as seen by
the various Hosts.  To make these things useable, you need, as a rule, to
have a file system.  File systems, also as a rule, do not appreciate their
Volumes coming and going.  They also do not share well with other File
Systems.  (A little matter of locking and Meta data makes this difficult.)
So the only real possibility here is for exactly the same File System on
exactly the same homogeneous host to be started serially on various
different systems, when they know the other systems have stopped and
flushed all their caches, etc.  This is very unwieldy.  The other approach
is, that they must have a share file system, and that is a big deal, and
there are only a few in the world, and even fewer that might work across an
internet connection.

So based on the above, I for one, reject your "requirement" statement, as
an unproven iSCSI need.

.
.
.
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 10/09/2000 10:04:19 AM

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



John,

> 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.

In a normal drive environment, mounting information is rather static as is
the interconnect to storage.  Providing such an extensible SAN, one should
consider mounting dynamic drives.  In real life use, it would be extremely
short sighted to ignore this *required* mounting information.  Rather than
have the client store this information, the bootstrapping database can
greatly assist the client by storing mounting information to assist in
creating a suitable environment for running various applications.
Eventually, tools should allow the user to redefine these values.

LUN may be used at a higher level than the device.  As SCSI allows for this
resolution, the naming scheme for mounting should also include this
resolution.  Regardless in how a LUN is seen by the device, until such
addressing is totally removed from SCSI, and I doubt it will be, providing
needed structures for mounting based on user requirements should be
considered within the bootstrapping database at the LUN level.  A symbolic
means of organizing these drives is appropriate at this level of
configuration even if the LUN may always have the value 0.

Rather than burden transports with these details, acceptance of an external
authentication scheme removes much of the burden this transport must bear.
In the end, you will be developing an LDAP server within the transport
protocol from the direction of this present scheme, otherwise.  DNS does
not
provide the needed selectors found with LDAP.

Doug