It is a bit nebulous. What kind of benchmark do you want?
Simple RMI calls (without any business logic) to SLSBs are >1000/sec with one client.
But every other benchmark depends extreme to your application.
One plus is the clustering feature (EJB) to be scalable.
So if you ask with a bit more detail and background what your looking it is easier to give an answer.
Thank you for your reply.
I am particularly interested in a collection of benchmarks regarding EJBs and clustering that can be found on a web page.
>>Simple RMI calls (without any business logic) to SLSBs are >1000/sec with one client.
Please may I know where I can find the reference to this benchamrk info you had mentioned.
Both Unisys and HP at one time or another worked on benchmarking JBoss AS (though for earlier releases than 5.x) using SPECjAppServer. Each presented some of their observations and tuning hints at various JBoss World conferences. However, the SPEC license agreement specifically prohibits publishing any SPEC benchmark result without approval from SPEC, which means that any official benchamrk results would have to be found on the SPEC web site. Last time I looked on their web site there were no JBoss AS numbers posted.
The test are all self made ;-)
I use it to have a simple benchmark to evaluate the different HW plattforms, cluster, JVM and JBoss versions perform.
I use a simple (configurable) mulltithread JavaApp and a simple SLSB to check RMI
and a simple servlet with JMeter for HTTP.
For our application we have (still in progress) a test suite to check the performance accross the different application version.