|Monitoring Server Performance||KEVIN RESCHENBERG|
There is an apparently little-known utility available at
PeopleTools releases 8.19 and up. "PeopleSoft Ping" will
attempt to time trips to the web, application and database
servers and then report the results. If your system suddenly
starts performing slowly, you might want to try it.
One reason that many people don't know about this utility
is that it is unavailable to users by default. You must
grant access to it through security. It is located on the
PeopleTools | Utilities menu.
The Ping utility consists of a page that refreshes itself
as often as you specify. The results of each trial are saved
and are available with a delivered query called PING_PAGE_RESULTS.
One client of mine was approaching a benefits open enrollment period
on a new architecture and wanted to monitor the system's
performance closely. I threw together a quick-and-dirty
page of bar charts that were created (in plain HTML format)
by PeopleCode based on the results of Ping. It worked very
well. Once it was done, I started it running continually...and
it crashed. For some reason, Ping might run for 15 or 30 minutes
or some other random time period but would then stop with
a page error. I never discovered the reason, and fortunately
there were no performance problems with the open enrollment.
(We just watched the servers with
the Windows Task Manager through Terminal Services—a
simple alternative if your server machines are running Windows.)
Don't try to
interpret the numbers on the Ping page in absolute terms, but
instead watch how they change relative to each other over
time. The numbers don't mean much by themselves, but they should be able
to tell you where a problem has developed.
At 8.4, Ping has been greatly extended and includes delivered charts. The 8.19
version could be considered a first step in that direction.
Now, I need to point out that if you have any of the good
third-party tools for monitoring servers, you probably won't want
to use PeopleSoft Ping. And of course it adds its own overhead,
so if your system is very slow during a critical time, running it
is not going to make anything run faster. But with that in mind,
you might want to check it out to see if it gives you any
insight into bottlenecks in your architecture.