UA in large address spaces

scheible at vnet.ibm.com scheible at vnet.ibm.com
Thu Jul 14 08:58:40 PDT 1994


   Regarding Tom's note attached below, I thought it might help if I
let you know how SSA is handling this.
   Since SSA also has a 8 byte address space, it is not practical to
maintain initiator information for 2 to the 64th possible initators.
Therefore, SSA has a registration process (Query_node/Query_node_reply)
during which the nodes create an initiator table.  This table is
limited, and the Query_node_reply message has a bit ITF (Initator Table
Full) that basically rejects any new initiators once the table is full.

I hope this helps,
John Scheible
IBM
Bldg 815,  Mailstop 4051
11400 Burnet Road
Austin TX, 75758
Voice: (512) 823-8208
FAX:   (512) 823-0758
EMail: scheible at vnet.ibm.com
=======================================================================
Date: Wed, 13 Jul 94 16:31:28 MDT
From: Tom Wicklund <wicklund at intellistor.com>
Message-Id: <9407132231.AA20983 at ebisu.ISI.systems>
To: scsi at WichitaKS.NCR.COM
Subject: Unit attention and SAM

The SCSI Architecture Model describes Unit Attention conditions
(section 4.6.5) as being kept on a per initiator, per logical unit
basis, as SCSI-2 defines Unit Attentions.

How is this reconciled with the advent of 64 bit initiator
identifiers?  It appears that a protocol which uses an N bit address
(e.g. FCP appears to use a 24 bit address) requires that the target
keep 2**N bits for each Unit Attention condition possible.  For power
on reset alone, this implies that an FCP device must have 2MB of
memory just to keep power on reset bits in order to comply with SCSI.

Is there some sort of limitation on the number of initiators a device
can talk to between power cycles or other method to limit this?  Or
are SCSI-3 devices required to keep per-initiator unit attention
status for all possible devices?

Thomas Wicklund
Fujitsu Computer Products of America




More information about the T10 mailing list