Multi-port vs. Multi-target SCSI devices

Edward A. Gardner eag at ophidian.com
Thu Jun 29 16:20:30 PDT 2000


* From the T10 Reflector (t10 at t10.org), posted by:
* "Edward A. Gardner" <eag at ophidian.com>
*
>* Gene_Milligan at seagate.com
>*
>But if instead of jumping to an assumption that it is one LU and that it is
>Fibre Channel instead derive how many LUs there are with two parallel SCSI
>ports and one media (perhaps I should say device server but at the moment I
>have comments on the use of device server in SPC).

I used FC and a single LUN for illustration.  The same argument holds with
other protocols and multiple LUNs, namely that there is negligible
difference between one multi-ported Target and multiple single-ported
Targets.  If you have a counterexample, please give it (that was the point
of sending this to the reflector :-).

To give another example, suppose we had a large RAID (excuse me, I mean
SCC-2) device with two SPI-3 "ports".  I contend we should describe this as
two distinct Targets, each with its own Task Manager, each with its own
Target Identifier (aka parallel SCSI ID).  Each LUN of course has its own
Device Server which, since it is associated with the LUN, is common to or
accessed by both Targets.

The alternate model, that we have a single Target with multiple ports, adds
a lot of complication with negligible (if any) difference in behavior,
therefore at best negligible benefit.  It should be removed from SAM-2.  I
think it is already absent from the other SCSI standards.

Note that I would probably still describe this as a multi-port SCSI device
in product data sheets and the like.  It's just that in formal SAM-2 terms
each port has its own Target entity with its own Target Identifier (aka
parallel SCSI ID).  If both ports happened to be attached to the same SPI
cable, then a single host adapter would see them as two distinct Targets at
two distinct IDs (and there would be two distinct I_T nexuses possible).  I
further contend this is already the way such devices actually work, with the
minor detail of exact TARGET RESET semantics.

Edward A. Gardner               eag at ophidian.com
Ophidian Designs                719 593-8866 voice
1262 Hofstead Terrace           719 593-8989 fax
Colorado Springs, CO  80907     719 210-7200 cell
-----Original Message-----
From: Gene_Milligan at seagate.com <Gene_Milligan at seagate.com>
To: t10 at t10.org <t10 at t10.org>
Date: Thursday, June 29, 2000 4:28 PM
Subject: Re: Multi-port vs. Multi-target SCSI devices


>* From the T10 Reflector (t10 at t10.org), posted by:
>* Gene_Milligan at seagate.com
>*
>But if instead of jumping to an assumption that it is one LU and that it is
>Fibre Channel instead derive how many LUs there are with two parallel SCSI
>ports and one media (perhaps I should say device server but at the moment I
>have comments on the use of device server in SPC).
>
>Gene
>
>
>*
>* For T10 Reflector information, send a message with
>* 'info t10' (no quotes) in the message body to majordomo at t10.org
>

*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list