Re: flaky IOC problems at Jefferson Lab

[wbrown@csg.lbl.gov (Bill Brown)]

Hi, Marty!  You said:

> No matter what improvements are made to the CA connection algorithm,
> it will not solve the problem of an IOC having little or no idle time.
 

Alas!  This is always true, no matter how hard we wish it wasn't!  It's
amazing how often we go down this path.

We got bit on a similar sort of problem on our system that talks to the old
control system via a Bit3 vme/multibus-I interface.  Four mv167/8 Meg
boards couldn't handle the load.  (The symptoms were somewhat different,
but we hadn't mucked with priorities.)  We replaced 3 of them with Heurikon
68060/16 Meg boards and the problems went away.  Oddy enough the Heurikon
boards are less expensive than the smaller/slower '167 boards.  We didn't
replace the the 4th cpu because it's the system controller, and Heurikon
admits that their sysCon section has its problems.  And the ca load on that
particular processor is not too high.

Sometimes we forget that our time is not "free."  It doesn't take many man-
hours to burn up the cost of a new processor or three.  OTOH, like all
engineering efforts, there are practical limits and the ever-present "Law
of Diminishing Returns."

For some reason, I seem to like to state the obvious on Tuesdays.

Disclaimer:  Any opinions are my own and have	    |  -bill
    nothing to do with the official policy or the   |   wlbrown@lbl.gov
    management of L.B.N.L, who probably couldn't    |   Berkeley, CA
    care less about employees who play with trains. |   aka wlbrown@netcom.com