which app server are you using now? does the jboss improved its performance over the last 2 years or have you switched to Orion or another app server?
We started developing an application using JBoss/Tomcat, and almost immediately noticed that it was slow, both through Tomcat, and through accessing the EJB directly for the Unit Tests.
After a certain level of functionality was put in place, we did some load testing, and found that we could only handle a load of 15 users. After some profiling, we boosted that to 20. Using Resin instead of Tomcat would only boost that to 25. That's still awfully low. We have tuned updates on, logs set to minimal, and we don't seem to be having RAM issues.
Preliminary tests with Orion indicate that it could handle well over 500 users using the exact same application, in less RAM than JBoss /or/ Tomcat /or/ Resin.
Unfortunately for me, I've always had issues with the support and documentation that follows Orion about, and JBoss has a number of really well-implemented features (the verifier and deployer are very nice, the error messages are usually relatively clear, and I can look at the source if something goes wrong and no-one can help me).
So, a couple of questions:
Why on earth is JBoss so slow, and is there anything else we can do to increase its speed? I can't guarantee we don't have some inefficiencies in the code still, but 20 users is awfully low even for inefficient code. Also, Orion doesn't seem to suffer from the same problem.
Is there anything we can do to increase the speed?
Do other people have similar issues with JBoss performance?
Not trying to knock JBoss here - I've got a lot of good things to say about it, but if I can't get more speed out of it than I currently am, then there's just no point in using it.
Being able to handle only 20 concurrent users is not normal. It's however impossible to say where your implementation is going wrong without seeing the code. I'd suggest getting a Dev. Support and have someone look over the implementation to see where it goes wrong or if there's a configuration mistake somewhere.
I see you dug up a post from year 2001, there's an entry in the Performance Tuning forum where someone said their app on out-of-the-box JBoss performed better than out-of-the-box Orion. Not that these mean much since each deployed application behaves a little differently so you better do your own testing.