VAX 11/750 rescued, alas...
shoppa_classiccmp at trailing-edge.com
Tue Oct 9 04:59:52 CDT 2007
"Ethan Dicks" <ethan.dicks at gmail.com> wrote:
> I know I went over bits of this before, but perhaps a different
> approach might be clearer... Systems Industries made a variety of
> disk/tape systems for various DEC machines. I have personally seen
> Qbus and 11/750 host cards, and I think there were others (Unibus, at
> least). To order an SI system, you'd tell the salesman what box you
> had, and what devices you wanted to attach. They would spec out the
> right host controller, and the right cards for the SI9900 external box
> to do what you wanted to do. I think there was also support for
> multiple hosts to access multiple disks - not like a true cluster, but
> more like multi-port access between a set of CPUs and a set of disks
> and/or tape.
The SI9900's supported (with the right options) multi-port access
over the SI-proprietary interface, and could also be used with
two SI9900's (one for each CPU) and dual-ported SMD drives. The latter
is what I more commonly saw.
And it caused no end of heartache! A firmware variation between the
two SI9900's would render a disk written by one of them, impossible
to read on the other. Almost as bad as firmware variations in modern
RAID controllers, in that it inspires you to keep so many identical
ones around "just in case" because you know that field service will
not be able to find the right firmware when you need it.
> When I used an SI9900, it was nearly as simple as it could be. We had
> an SI 9700 board in our 11/750, a pair of 40-pin cables to SI 9900 box
> in the next rack, one host-side card in the SI 9900, and two SMD disk
> cards in the other half of the SI 9900 box. We had two SMD disks
> (60-pin control, 26-pin analog data) in the rack with the SI 9900,
> emulating a pair of RM03s and a large RM05. We could have chosen to
> hang more disks off our SI 9900, but the way SDI disks were plunging
> in price in the late 1980s, we kept the SI 9900 for our boot disk, and
> added a couple of RA81s for user data, etc.
The Fuji SMD drives plunged in price in the mid 80's, and I don't
think SDI ever got that inexpensive :-).
> There were some Emulex Unibus SMD interfaces, but ISTR those were more
> common on PDP-11s and VAXen running UNIX, not VMS.
Under VMS the Massbus-emulating interfaces were a pain in the butt
if you used a non-standard geometry, because you had to patch all
the drivers in all the right places. Still lots of sites had them with
VMS. After MSCP emulation came in it became so much easier!
More information about the cctalk