[bit.listserv.vmxa-l] A tuning experiment

TYJ@SLACVM.BITNET (Ted Johnston) (02/28/90)

I read your comment on your exec for setting tuning parameters and
I am interested to know what you use to determine if you are:
   CPU bound
   Storage
   Paging
We have used an automated approach for about 1/2 dozen years., but
we have been more simplistic - just trying to keep expansion factor
down because of excessive CPU utilization.  It worked fairly well
under HPO on our 3081 with solid state paging devices.  On our
3090 under HPO with XSTORE expansion no longer was a good measure
(it frequently greatly over estimated expansion).  Under XA
I have never seen expansion go above 2 for Q3.  We are now using
the number of users in Q3 divided by 2 (we have a 200) minus 1.
This number comes close to the number obtained by a REXX exec
that is compute bound and calculates its CPU absorption versus
elapsed time.  Ted Johnston
               SLAC