I am planning to upgrade my web server (LAMP) which is having a
E3-1271 v3 @ 3.6 GHz processor (4 cores), with hyper threading enabled. The current cpu utilization average between around 35% – 45%. I read that with hyperthreading enabled, the cpu performance takes a hit above 33% utilization.
I have also read that processor intensive single threaded applications will benefit from higher clock speed, whereas web servers will see better performance with more cores. With higher clock speed, threads will be pre-empted less often.
Is there a metric that shows how often the processor preempt threads, to give cpu time to other threads. Is load average a good metric for determining this?