iSCSI URL scheme

[Black_David@emc.com]

With my co-chair hat off ... I've had a long-running concern about the URL
naming in iSCSI that Daniel's example allows me to express concisely ...

> Let's imagine a large array of fast, reliable and featuresome storage---an
> IBM Shark box say.  B-) This box has many user accounts, and is being used
> in the context of a storage service provider.  The login sequence might go
> something like this....
> 
> I: Hello Shark---I'm an initiator.  My authentication ID is "Robert Smith,
>    Seattle, next to the donut store, hjhhaskjanbt112kg9d988".
> 
> S: Hello initiator.  I'll be your server---we need a secure connection:
>    switch to "hjcnneamchjsdkfsdsd394njsdf89423".  (Trivial example.)
>
> I: <encrypted> Alright!  I want to connect to
>    "scsi://niceshark.acme.com/bobsstorage/san3?wwn=0x7b4d72658a7410db"

This query is a *major* departure from SCSI practice at this level.
The SCSI approach would be more along the lines of:

I: <encrypted> Alright! Show me all your LUNs that I can access.

(i.e., REPORT LUNS) at which point the Initiator then does additional SCSI
commands to figure out what each LUN actually is.  One of these commands
could
be to read (or attempt to read) the WWN for each LUN.  For disk storage that
is
under a volume manager, there's another round of discovery later on in which
the volume manager reads the volume label and figures out what the storage
is
independent of any information that was used to discover it.

Returning to the three examples:

> Then we have a choice...
> 
> S: <encrypted> Great---off you go and have a nice day.
> 
> or
> 
> S: <encrypted> Sorry---you haven't got permission to use that any more/it
>    doesn't exist.

This is an authorization check that can be implemented by a combination
of accepting/denying the connection based on the client authentication
information and/or only reporting the LUNs to which the Initiator has access
(and after an exhaustive search, the Initiator may discover that a LUN it
used to have access to [based on WWN} doesn't seem to be there any more).
Fibre Channel makes widespread use of a mechanism similar to the latter at
the
target discovery level (nameserver based soft zoning, where the fabric
nameserver
only tells an Initiator about Targets it can access, rather than all the
Targets it
knows about).

> or (just after they buy more storage)
> 
> S: <encrypted> Ok, that URL has been changed to
>   "scsi://bobstorage.san3.acme.com:9003/".  Please come again.

This is a common computer science solution to a large set of problems;
introduce another level of indirection ;-).  While I understand the
solution,
I'm not at all clear on what the problem is and why it needs to be solved
in this fashion.  There's been a lot of traffic on this, have I missed
something?

The counterbalance on the other side of this issue is that iSCSI
should fit into existing SCSI frameworks for storage discovery,
access, and management (e.g., in operating systems).  Every time a
new sort of naming is introduced, that integration difficulty goes up
considerably.  The worked example is that Fibre Channel had to introduce
WWNs or their equivalent for some fairly obvious functional reasons,
but introducing WWNs to operating systems that started out only
understanding parallel SCSI caused no end of pain and suffering.

Something that's going to rear its ugly head here is discovery.  SCSI
discovery is fundamentally based on a "bus walker" paradigm in which
low level logic does discovery based on a comprehensive search
(i.e., try every possible target on the parallel bus).  When Fibre
Channel introduced WWNs, this broke the "try every possible
target" approach (can't iterate through every 64 bit WWN and still
boot quickly), and it was fixed by a multi-round distributed address
assignment scheme for FC-AL (don't ask ;-) ) and the ability to
query the nameserver for all known ports in FC-SW.  FWIW, FC-SW
tried to use LDAP for nameservice and gave up; something considerably
simpler was implemented.  Among the practical problems that resulted
from WWN introduction is that booting over Fibre Channel took a while to
get working because FC broke the typical "first LUN of first target contains
a boot image" mechanism used in boot code.

iSCSI has the opportunity to repeat the Fibre Channel experience of
changing the way things work in a fashion that will take a while to
work its way through systems.  If this is done, it needs to be done
for a good reason.  One question to think about is "If all SCSI targets
are named by URLs on the server, where do the URLs come from?"
An answer of the form "yet another config file or set of registry entries"
is an obstacle to adoption, because tools have to be written to manage
these, and administrators have to learn how to use them, and ...

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