"galder.zamarreno(a)jboss.com" wrote : | Effectively, this means that GC has it. Are these machines multi core? What are the startup options you're using? Did you enable GC logging? Prabhat, I thought we went through similar situation when you're testing 1 node scenarios, didn't we? I think I suggested that you use tools like TDA to discover scenarios like this. Yes, we went through that exercise and because of that we are at this stage. Portal was tuned enough to get the desired performance in one node and that is taken as baseline for scalability testing. As you can see there is no scalability problem with 3 nodes and now with 4 nodes. So, I do not know why GC would be a factor in 5-node cluster when it's not in {1,2,3,4} cluster. Just an FYI: all these machines are of same architecture (multicore 64 bit) and are in the same subnet. Please ping me as needed with server setup information. I did see few threads waiting on DB connection which Brian also pointed out. This was intentionally left out because caching whatever data that threads are trying to get would result into a huge memory footpring in some use cases. But what I am going to do is turn on caching for those data and see what that results into. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4170108#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...