Integration

Hello,

We are currently running load tests against our Coherence cluster. We are connecting via a proxy service which fronts the actual instances. We are currently running into a large performance hit when we ramp up to 300 users (I know, the number is arbitrary/meaningless without context). We have 2 application servers that are clustered (2 instances of our web app) and they are talking to a cluster of coherence servers also on 2 seperate boxes. All boxes are multi-processor and multi-core. One of the settings that I'm curious about tweaking is the thread-count element. Both on the client and the proxy/service instances. I've done some googling, and some people say this value should reflect the maximum number of concurrent threads attempting to access the service while other tie it the number of physical cores/processors. Can someone provide some insight into how one should determine that value?

Thanks!