I was recently involved with the installation of an IBM 7133 Disc Subsystem
solution, currently shared between two HP3000's; one 980/100 and one 947.

During initial 'proving' the 7133's were connected, via a Vicom
Ultralink2000 adapter, to the 947 via SE-SCSI. Which worked well.
The reason for SE-SCSI was that the 980 system HAS to use SE-SCSI.

Subsequently, it was decided to use FW-SCSI to connect to the 947 in order
to obtain the best possible performance. That is when the problems started.

In the end, it was deduced that the problems encountered were due to MPE
sending invalid data on two bits beyond the 8 bits utilized by the SCSI-2
standard - which was causing the 7133 disc drives to hang.
The problem was resolved by Vicom modifying the microcode in the Ultralink
adapter, effectively blanking out the offending spurious data from MPE.

It may be that few sites, if any, are also utilizing this configuration.
However, it could be that the invalid states being transmitted by MPE
through FW-SCSI may be having a detrimental effect on 'other'
configuration's performance - without actually hanging the drives.

The feedback that we got from Vicom is that HP are not going to fix the
problem - but, why should they?

If anyone else has had similar problems, or if anyone at CSY has any
comments or inputs, I would be pleased to hear from you.

Robert.