In order to determine what is good performance you need understand how much time it takes to get to the class through the EJB. I would like to know:
1. How are you invoking the bean via a web service or using RMI?
2. What happens if you run the test case via the EJB several times versus just once?
3. What is the number of EJB's that are instantiated prior to calling JBoss?
My 2cents says if you have simple business logic in your class regardless of who's container you use JBoss, WebLogic, WebSphere the class will always run much faster because you are not going through all those layers in a container to get to the class.
Hope that helps.
1. The bean is being invoked via RMI.
2. If the test case is ran several times, there still isn't much difference in overhead.
3. As far as I know, there is a call from a Stuts Action class to a thread, the thread then invokes the EJB, the EJB then makes the call to the business bean.
As for the performance issue due to going through all the layers, I appreciate there will be a greater overhead, but I am taking performance timings for just the class itself - not how long it takes to get to it. So layers or not, it still takes longer for the classs to execute while inside the container.