RE: Detecting execution bottlenecks

This might sound a little too simplistic, but it is often a good way to
start.

Put in a logging routine to log a session ID, application point and
timestamp to a file.  Be liberal with your logging, but be sure to focus on
points-of-transfer, such as calls to the server.  Run your application in a
dedicated mode, with as little load on the server as [possible and only a
single client task.  Do this a few dozen times to get some good benchmarks
for timings.

Then ramp up the load and run the same thing.  Make sure your session IDs
are clear, so that if you use multiple sessions from one client that you can
differentiate between the sessions.  Check your timings and find out which
gaps are taking longer.  This should give you a place to look for your
bottleneck.  Not 100%, because lots of things factor in, but it's a starting
point.

Joe

From: James H H Lampert

I've been asked to research an execution bottleneck in one
of our client-server products. So far, I'm not entirely
sure where to begin, except that for now, we're assuming
the bottleneck is happening on the server.

Can anybody suggest a good way to track down bottlenecks
in a client-server environment?