RE: iSCSI: Session Partial Resolution

[Black_David@emc.com]

Charles Monia asked me a couple off off-line questions whose answers
are of general interest:

> As I interpret the above, you are saying that there is a rough consensus
in
> favor of the asynchronous model  -- ie. one TCP connection to handle all
> control traffic. As an option, additional connections may be created to
> carry data traffic. Is that correct?

No.  The rough consensus is only for whether to put the command numbers
and sliding window information in the headers in the single connection case.
Both the symmetric and asymmetric model could be negotiated up from
there.

> I thought that that the command reference number and sliding window in the
> control PDUs was needed to make the symmetrical session model work.  As I
> recall, these allowed commands arriving on different TCP connections to be
> deskewed by iSCSI and sequenced for in-order delivery to the ULP.  If
these
> are no longer present in the iSCSI header, is any meaningful consideration
> of the symmetrical model possible?

They could be added by negotiation at the start of the session, but it does
raise the bar for consideration of the Symmetric model.

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