Howard,
One thing you didn't say -- and maybe you don't know -- is whether the hit
percent you are seeing is new. A hit rate in the 90's is not unusual for
MPE systems, but the question is whether it is typical for your system. If
you have experienced a sudden degradation, then you should look at recent
changes, such as how many and what kind of jobs are running concurrently,
any new SUPRTOOL procedures recently implemented, any significant database
changes, etc.
Whether more memory would help your situation depends on the cause. Memory
pressure might be a significant factor. If so, you would probably be seeing
a lot of Memory Manager overhead and a high page fault rate. A more likely
problem is poor data locality. Do you do any regular database and disc
maintenance, such as defragmentation and detail set repacks? If it's been
awhile since you did those things, you could see some significant
improvement by attending to them now. (I know - it's hard to get any system
downtime to perform maintenance...) In some cases, a redesign of some of
your database queries might be needed to achieve any significant
improvement.
-----Original Message-----
From: Hoxsie, Howard [mailto:[log in to unmask]]
Sent: Wednesday, July 17, 2002 3:05 PM
To: [log in to unmask]
Subject: Memory Utilization Performance Question
Hello All,
Here's one from everyone's favorite problem area, performance evaluation.
We have both Glance and SOS on our 989-650 (6.5pp2 w/May2002-reactives) and
when the help-desk folks started reporting poor performance complaints from
the users today I started taking a much closer look with both utilities. I
understand that Memory Read Hit % should normally be seen in the mid 90s %
whereas ours has ranged between 58% and 72% over a 1 hour period.
I have been banging my head against this issue for quite a while and am
having a hard time pinning down the real cause of the problem (yeah, looking
for that magic bullet again--I keep forgetting where I keep it). Has anyone
else experienced these symptoms before? In other words, do any of you have
any magic bullets you can share with me?
We have 3.7 Gig of memory (we still have a PSI card so we can't get to 8 gig
until we remove it or add on a beta patch). We run Ecometry software. We
do 10 times (according to a highly regarded performance specialist we
contracted for a performance study last year) the number of SUPRTOOL
routines during each week than a normal system. We have half a dozen jumbo,
waaaaaay-jumbo, datasets in our environment. Last month we replaced our old
disk array with a new one with disks nearly 3 times as fast as the old one
and some batch processes really improved while others stayed the same.
ANY ideas, hints, suggestions, recommendations, or offers of lucrative early
retirement will be graciously and gratefully accepted.
Howard Hoxsie
Sr. HPe3000 Systems Administrator
Nordstrom-direct, inc.
www.nordstrom.com
(206) 215-7069 voice
(206) 215-7869 fax
(206) 356-0683 cellular
[log in to unmask]
600 University Street, Suite 600
Seattle, WA 98101-4102
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
