I am doing some performance tests to determine response time and the ability to handle concurrency with various cpu / ram / os configurations. One interesting find that I came across is that it seems that one jvm works better with 4 cores than with two cores (not surprisingly), but adding more cores for 4 cores does not lead to any significant improvements. But then adding another jvm instance with a load balancer (same hardware) led to a significant improvement.
It seems that an individual process is limited by the number of cores that it can use, possibly due to the limitation on the number of os threads that the process can appear at the same time. This is a 64-bit environment.
Im using tomcat and tried to change the maxThreads attribute, but this did not affect the amount of Im concurrency that needs to be processed.
Any other reasons for this can be explained?
source
share