[sjtrny]

You wouldn't need to monitor individual threads and processes. Only average response time for each server or total CPU load.