[comp.periphs] VMEbus--UNIBUS conversion

rayan@ai.toronto.edu (01/29/87)

We have 3 CDA array processors sitting on a VAX-11/780. I hope eventually
to replace the VAX with Suns of some ilk. However, in order to do that, it
is necessary to find some way to retain the functionality of the CDAs; we
use them primarily as display devices. To do that, there are two possibilities:

1. Sell the CDAs and buy something VMEbus compatible.

2. Find a way of hooking the UNIBUS-only CDAs to a VMEbus.

Item 1 is not likely to be practical without a large infusion of funds from
somewhere... unless of course someone who reads this would like to buy them
off of us. Comes with lots of cross-development tools. Of course, if somebody
would give us a Grinnell or a Pixar, that would solve the problem. I digress...

Item 2 is the object of this query. Since the CDA is a dumb device, and
since hopefully the '780 will go away at some point, we are left with a
bunch of relatively dumb cards sitting on a UNIBUS backplane, without a
processor to keep them company. The problem is how to get them to talk to
a Sun & vice-versa.

We have looked into the following approaches:

- Dig up a busconverter. The company that everyone seems to think is making
  VME-UNIBUS converters doesn't, and basically told us to not count on them
  for such an item. If you know of someone who makes one or is about to bring
  one into production, please let me know!

- Dig up a UNIBUS-to-any-other-bus-that-there-is-a-VMEbus-converter-for.
  We know of no proven solution. If you do, please please let me know!
  Daisy-chaining bus converters is likely to be the cause of wonderful
  problems....

- Put a DR11W on the UNIBUS and use standard VMEbus--DR11W interface.
  This won't work without coding and burning new proms for the CDAs, since
  the DR11W is just as dumb a device as any other and needs a processor
  to make proper use of. Theoretically, the CDAs could be reprogrammed to
  poke it appropriately, but we don't want to do that.

- Get an El-Cheapo PDP-11 to stick them onto. The change from the present
  environment (4.xBSD) on the VAX to the PDP would be just as traumatic
  as changing to a Sun. The OS would be eminently non-standard in our
  environment. This seems like a good idea for the hardware, but a bad
  idea for the software.

- Get an El-Cheapo VAX-11 to stick them onto. Something like a '730 perhaps.
  We can run sane UNIX on it, and no tools need to be ported or changed. Is
  the best solution from a hardware/software point of view. The possible
  problem here is economics. How little are people willing to let their
  '730s go for? The market for these things isn't exactly overwhelming,
  so I would expect them to be quite cheap. All that is needed is a barebones
  system -- we can always add in some distributed file system if we need
  gobs of disk space.

- Get a not-so-El-Cheapo microVAX-II to stick them onto, along with a
  Qbus-UNIBUS converter. Fine solution, except a bit too expensive unless
  we can get a used microvax for the same price as a '730...
  I gather that the Qbus-UNIBUS mapping in uVAX-I's isn't transparent, which
  seems to make them less desirable for the purpose. Is this correct?

- Build a busconverter in-house. Not infeasible technically, but we have
  better things to do with people's time. If anyone has done this already,
  I'd like to hear.

Questions:

1. There is this crazy idea of using a VMEbus-Qbus converter, and hook it
   to a Qbus to UNIBUS converter. Comments on the practicality of this idea?
   [ I can hear people fainting... ] I am not sure if the VMEbus-Qbus
   converter is transparent or uses (sigh) a DR11W interface. Anyone know?

2. Anybody know of a UNIBUS-to-any-other-bus converter that does NOT use
   a DR11W, and which might eventually be connected to a VMEbus?

3. If you have a 730 (or uVAX-II) taking up valuable rack space, what would
   you be willing to sell it for (I would just like an idea of what we would
   have to pay to get such an item, but if you want to more or less give one
   away now, I'm listening)?

4. Are the prospects of selling off the CDAs as bleak as we think?

5. Any other approaches we haven't thought about?

One would think there is a large market for VME-UNIBUS converters. How come
no one seems to be making one? If VME-Qbus converters exist, what is the
big difference to explain the lack of UNIBUS ditto?

Any and all help would be appreciated.

rayan