We are running a few WebDynpros (ESS Travel & Expenses and one custom Java Dynpro) as well as some PCD content in a production environment. We have a huge amount of resources available to run this: four servers running 5 JVM's with 2GB of memory available to each -- and we are still running out of memory when traffic gets heavy! And by "heavy," I mean only about 450 active sessions for half that many users.
This is completely different than I am used to; we run almost 20 other high-load applications in a "regular" Java environment with just two JVM's at 160MB each. So my questions are:
- What can we do to reduce resource consumption? Any ideas are appreciated. Is there a magic parameter I should look at?
- Is this resource consumption consistent with what you observe on your Portal?
- I have a gut feeling that it is ESS Travel & Expenses that is the primary resource user. Does anyone else running T&E see this behavior?
- Is there a particular report I can view to see which Dynpro is the main resource user?
- Last time I analyzed a heap dump, it looked like ApplicationSessions were the culprit. Lowering the timeout value for WebDynpro sessions helped a lot, but not enough. Is there a way to reduce the resources consumed by a user's session?
Your input is much appreciated!!!