It has been a while since I played with JMetere, but I recall that it has a mechanism to test EJBs directly, and not through a web page (ie, via http). You are not, by any chance, using a web service to access the EJB? I assume you have JMeter accessing a servlet or JSP which in turn is accessing the EJB. But my original assertion still holds: anonymous wrote : More is not always better. The more threads you allocate, the more contention there will be for the CPU(s). There is always a sweet spot when setting any pool size. Going either above or below it causes performance degradation. The key is finding the sweet spot, which usually involves repeatedly running the test using different pool sizes. Of course, you also have to determine if the pool size is what is causing the bottleneck - if the bottleneck is elsewhere in the system then no amount of pool size tweaking will help. If you don't know how to do this yourself, you will have to hire someone to do it. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4137607#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...