RE: iSCSI gateways, proxies, etc.

["Douglas Otis" <dotis@sanlight.net>]

Jim,

Many of the typical functions offered by SCSI for discovery do not scale
should there be significant numbers.  The complexity of device 'views' is
only a small aspect that can not be practically solved within the transport
itself.  Once a conclusion is reached on the need of an external database to
handle these many complexities driven by aspects of authentication, device
addressing, mounting information, permissions, as well as WWN,
virtualization of the encapsulated Extended Copy address looks like a good
idea.  With a user view within the database and a far more extensive server
view, responsibility of verifying permissions together with any redirection
is placed upon the SCSI access server and not the client.  The need for
mixing IP and SCSI addressing within the SCSI encapsulation transport
evaporate.  The device making the request should be able to assume a Copy
Manager can both verify permission and discover the path based on wisdom
granted by the database.  This process may be done by the access server
parsing and embedding what is actually needed.  This Copy Manger's path can
be as complex as one of David's nightmare double NATs but nowhere along the
way would a DNS be needed.

Doug


> Charles,
>
> In some sense I agree with you.  But I'd turn it around. In your scenario,
> it is MY OBLIGATION to give to A an identifier for B that is consistent
> with A's view of the world.  I shouldn't be so myopic to think that A's
> equals mine.   How I might discover A's view of the world is
> implementation
> dependent.  In fact, one might argue that if I don't know A's view of B,
> then I have no right to ask A to do something with the data on B!
>
> In your second paragraph, you suggest WWN.  I have to ask "whose WWN?"  If
> you mean the WWN of "logical unit" in question, then that is already
> covered (at least in EXTENDED COPY, see section 7.5.6.6 of SPC-2, rev 18).
> Unfortunately, the burden is then on A to find the logical unit with that
> WWN among all the possible locations (in the internet?) by exhaustive
> search (or dumb luck).  If you mean the WWN of the SCSI target
> device, then
> can you be more specific?   FC has WWNodeName and WWPortname, so that's
> cool.  Parallel has no such identifier (that I know of).   Some of the
> threads in iSCSI seem to be asking exactly the question of what
> that WWN is
> in the IP world.  In any case, it must be something that can be used by
> some well-known method to find an address for the device. (In FC, I can
> usually get from WWN to N_Port by clever queries to the nameserver, e.g.).
> DNS has this property (within a given domain), so it's not universally
> global.
>
> Jim Hafner