On Tue, 6 Feb 1996 at 09:09:44 -0600, Susan McClure said:

> Well IBM-  any ideas on how to trace/trap/debug this problem??  Looks
> like we have all been (mistakenly) keeping quiet, assuming it was "just
> us and our environment"

We are pretty well convinced that this problem is a symptom of the CPU
constraint on the server (it runs on only one processor of our 3090-600E)
and that the problem has nothing at all to do with our network.

What we have decided to do is start by proving or disproving this
hypothesis.

I intend to set up another server (on the same 3090) and move to it the
clients from the two user departments who are complaining most loudly
about the failures.  (This will be about 50 largish UNIX clients and
maybe some Macs.)

*Nothing* will be changed as far as the network.  That is, everything
between (and including) the client machine and the TCPIP service virtual
machine on the 3090 will be the same as before.  The ADSM server will be
the same release, etc., as we are currently running, but it will be less
CPU constrained.  (Our CPU constraint is not on the 3090 as a whole, but
only in the fact that ADSM can't use multiple processors simultaneously.)

I think this will clarify the situation.  Does anybody have any comments
or suggestions?

Melinda Varian
Princeton University