[list.netnws-l] Penn state 2.0 Netnews performance

Steve Moore <SWMGR%MIAMIU.bitnet@ugw.utcs.utoronto.ca> (02/13/90)

    If I have missed previous discussions of performance please let me
know.  However I would like to state some of the things I have done to
improve Netnews performance on a 4381-23 with VM/SP 5.0.
    Our site runs at 100% CPU utilization around the clock because of
large CMS batch jobs.  We have a solid state dasd so we can page at
around 700 pages/sec and still get subsecond interactive response time.
Our news feed gets us info for over 350 newsgroups and my netnews virtual
machine is unable to keep up with this volume when set at a priority
equal to batch processes but less than interactive users.
    A couple of weeks ago I set the priority of the Netnews machine
between the interactive and batch group.  My VMMAP reports showed that
for that week Netnews was using 15% of the CPU.  However we were able to
keep up with our news feed at this priority level.  Of course the batch
machine users were upset that their turn around times increased dramatical-
ly.
    Therefore, I would like to make the Netnews machine as efficient as
possible (as I am sure everyone else would).  So far I have Execupdted
and execloaded all the execs, increased the Netnews machine virtual storage
to 10 meg, and Nucxloaded Vlibcomp.  This lowered the average article
processing time by .15 sec.  Now I am going to order the IBM Rexx compiler
and look into some disk IO improvements.  I would be happy to spend some
time on improving the code (if possible).  I really want to see this
product fly.  Thanks to Linda and all the other contributors.  If you
you have any suggestions please RSVP.