Jorge, Threading helps get the most of your system by allowing a job to do work while other job is blocked for I/O, or by leveraging multiple CPUs. Because your tests do little I/O then the bottleneck will be the CPU. Increasing the number of threads in such a situation will actually decrease performance due to the overhead associated with context switching. Your tests do not necessarily prove that jBPM does not scale. They might just mean that your CPU is exhausted due to an overuse of threads. To prove jBPM is the bottleneck, you'd have to show that the CPU is noticeably below 100% usage and that there is contention between threads. Of course, it'd be interesting to dig further into this. Did you use a profiling tool that generated finer-grained data? Are your tests available publicly? View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4145034#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...