[1740] in linux-scsi channel archive
Re: SCSI IP
daemon@ATHENA.MIT.EDU (Randy Scott)
Tue Apr 22 18:47:47 1997
From: Randy Scott <scottr@belle.bork.com>
To: kragen@dnaco.net
Date: Tue, 22 Apr 1997 16:14:43 -0500 (CDT)
Cc: tech-kids@swcp.com, ben.elliston@compucat.com.au,
linux-scsi@vger.rutgers.edu
In-Reply-To: <Pine.SUN.3.95.970422160245.23556B-100000@kirk.dnaco.net> from "kragen" at Apr 22, 97 04:13:22 pm
> It should be noted that the prototype implementation described in the
> paper had a TCP throughput ranging from 150 KBps to 430KBps -- or 1.2Mbps
> to 3.4Mbps, a small fraction of the throughput available with Ethernet.
Note that there may have many factors involved with our lack of reasonable
performance. Now, I am primarily blaming the Adaptec AHA1520 host adapter,
which is known to be slow, and to asynchronous transfers. The maximum
360Mbps (or 400, or whatever) can only be achieved using Fast, Wide,
synchronous SCSI-II transfers.
> > What I'd be more interested in would be the *reverse* of this idea: SCSI
> > over IP, if you will. Go to IPv6, give every hard drive a network card,
We discussed working on SCSI over IP as an alternative project. However,
we felt that IP over SCSI seemed more interesting and possibly had more
relevance.
> > infancy. I think what you did sounds like a blast for a senior academic
> > project. But I fail to see where it has any practical application, today.
>
> I just thought it was megacool.
Definately a blast. I (not speaking for my partners) had a really fun time
playing with the stuff.
--
Randy Scott <scottr@belle.bork.com>
