Stateless session beans do no have identity so every concurrent request will actually be directed to a different instance from the stateless session bean pool. The server dynamically manages the pool size to deal with the request load.

I believe the multithreaded calls on the same handle will go to different bean instances.

I think that it depends on internal implementation of an application server. As I understand a SLSB should not has internal state. In this case, in theory, one instance of the SLSB enough to serve all client's call.

No. A bean is always guaranteed to be called by one thread at any given point.

"dimitris@jboss.org" wrote:
No. A bean is always guaranteed to be called by one thread at any given point.

What you miss is that SLSBs *can* have internal state, e.g. a cached datasource handle that they re-use to serve incoming requests.

Stateless is the interpretation from a client's point of view, so if a SLSB has 3 methods, a client should be able to call those methods on different bean instances, without that making any difference in the total outcome.

The single thread guarrantee of EJBs makes writing those beans easy, so you don't have to think about concurrency issues. It also scales well as the server will create as many SLBSs as they are needed to handle the load.

What you describe is a simple RMIServer singleton that must know how to deal with concurrent requests. That's a very different thing from EJBs.

If there are more than one threads from a single client, can they access the same SLSB simultaneously? If so, what exception would be thrown?
Thanks

The best thing to do is to make a testcase and see what happens.

I think it's thread safe because one bean is created for a particular client, and this bean just holds the data of this particular client, that means the data are not shared by other clients.

Yes it's thread-safe. For more details you can check the "Pro EJB 3" book.

Yes, your answer is correct. But about 7 years too late

You should have a look to the last post date

Grigoriy and Wolf-Dieter, thank you both. I didn't notice the data when this post started. Anyway your answers confirm my understanding.