>From: mxs46@po.CWRU.Edu (Michael Sokolov)

>   What I don't quite understand is why do they use unit plugs in the first
>place. SDI uses purely radial connections for multiple drives, and each
>drive connects to a different port on the controller. On all machines with
>RAxx disks that I have seen (only two really :-)), there was a perfect
>match between the number of the port on the controller and the number on
>the unit plug on the drive. Does anyone have a clue as to what's going on?

The docs encourage Field Service to install it that way so it's no surprise
that drive 0 is usually on port 0 etc.  However the MSCP spec requires that
unit numbers be unique on the *system*, not just within the controller;  so
once you add the second MSCP controller the port-unit matchups stop working.
Seems like an annoying rule though, all other types of disk/tape units only
have to be unique within the controller.

>   This looks really strange. A machine with 3 RA82s, two network
>connections, a lot of serial ports, and CI VAXcluster hardware is a very
>high-end configuration. 8 MB is just ridiculously small for such a
>configuration. If this indeed the case, and if you decide to use this
>machine, you'll have to spend some $$$ on memory.

Well of course it depends on the software.  RPI used to run an IBM mainframe
with literally hundreds of terminal users, a large laundromat worth of disks
(many GB),  half a dozen 9-track drives, an assortment of bizarre net
interfaces, and 16 MB of real memory, and it would just purr right along.
16 MB seemed like a lot at the time, and this is only 10 years ago.  I
would imagine VMS could do pretty well with that amount.

John Wilson
D Bit