Ultra SCSI
Jim McGrath
/G=Jim/S=McGrath/O=QMAILGW/PRMD=QUANTUM/ADMD=MCI/C=US/ at qntm.com
Wed Feb 16 18:22:42 PST 1994
Subject: Time: 5:01 PM
OFFICE MEMO Ultra SCSI Date: 2/16/94
I just got back from T11 and PCMCIA stuff (I hate meeting overlap!).
I would like to thank people for their lively responses on Ultra SCSI.
My objective for the SFF SSWG is simple - to define the timings
required for short haul (i.e. "direct connect") single ended 40 MB/s
wide SCSI. I would be willing to accept a cable length restriction
of 1 meter.
The application environment is anyone using a SCSI disk drive in the
future, but particularly people with a large number of drives per
bus, where bus bandwidth is important (e.g. servers). I fully anticipate
people to minimize the effective cable by directly plugging in the
drive or using short pigtail type connections. Most of the "cable"
would in practice be PCB tracing. I have customers today who would
buy this product.
Note that in my mind this is not a question of defining these timings -
they will be defined, by the market is no one else. Just as single
ended fast SCSI was not sanctioned in SCSI-2, but was done anyway,
reving up SCSI to 40 MB/s will be done anyway, and for the same
reason - it is cheap to do compared to the alternatives, easy for my
customer to transition to, and a compelling feature. If necessary we
will work privately with people on this topic (shades of ATAPI). SFF
seems to be a reasonable compromise - a public forum, but one where
fast progress can be made by focusing on the prime objective and not
being distracted by the need to serve a wider purpose. Specifically,
I am not interested in making 40 MB/s work for everyone, everywhere -
that is not what my customers need. I am not opposed to long term
work on that topic (in X3T10), nor to making reasonable provisions in
the SFF stuff to provide a transition to any future X3T10 stuff. But
my original goal still remains - I want agreement on something by
the summer, otherwise the market will take care of it on its own.
Jim
More information about the T10
mailing list