The processor load essentially indicates the number of processor cycles in which the processor had to do something, and not just twist the virtual thumbs.
So, if your request really does the real work (instead of waiting for I / O to disk), then you should expect that the load will be high while the work is done, because the processor has something to do.
What you need to look for ends with processor cycles, since this is the time when the response time starts to rise.
If your problem is that the request is so short that the profiler cannot show you what you need to see, then think about using an automated tool to ask it to process hundreds of thousands of requests. This should help.
source share