Dual Access to UDA50/RA Disks

[dan at msdc.UUCP]

This is just a word of warning for all of you out there that have
notions of dual-porting the RA series disks.  After much hassle with
our local DEC office we finally got our hands on the MSCP documentation
so we could fix the Unix driver to support RAs shared between two
processors (not file systems, mind you, just drives).  After wading
through an incredibly opaque and obfuscatory swamp of verbiage, we
questioned our marketing friends' statements that the drives were
dynamically dual ported.  After allowing for the fact we were somewhat
"uninitiated" to the marvels of DSA (and after threatening bodily harm)
we were allowed to talk with an engineer in Colorado who knew the real
story.

Without giving away any of the marvelous secrets we agreed not to
disclose, it turns out that the RA drives are indeed dynamically
sharable.  However, THE **@@#@* UDA50 DOES NOT AND WILL NOT SUPPORT
dynamic dual access to drives.  The only way it can be done requires
about 800ms to switch a drive.  Unless, of course, you shell out
another $150K for a redundant HSC50 configuration--THEN AND ONLY THEN
CAN YOU ACTUALLY DYNAMICALLY SHARE DRIVES BETWEEN PROCESSORS.  (In all
fairness, the static sharing works just fine--if a machine dies, you
can indeed switch the drive to the other machine.)

If you are interested in how we could be so badly misled, just whip out
your systems&options guide and look at the descriptions for the RM05
and RP07 (real dual access drives) and for the RA disks.  See if you
can tell that the UDA50 doesn't support dynamic dual access!  We sure
couldn't.  And neither could the local sales office.  (And anyone who
says "That's what you get for listening to a salesman" will meet with
an exceedingly unpleasant demise.)

Dan Forsyth	(...sb1!msdc!dan, ...allegra!gatech!msdc!dan)
Medical Systems Development Corporation