We have been going thru the Performance Tuning doc for Apache and experimenting with the Directives one by one. We enable/disable or vary the value in Directives then submit some traffic
via "ab" from another computer and then look at RAM/CPU/Webpagespeed impact. This is a dev box noone else accesses, we have a locally installed webpagetest. The activity on it is fairly pristine at the moment.
The ab command is: ab –n 200000 –c 400
Doing this process for "HostNameLookups" we got this …
The hostnamelookups result is a bit interesting. As the document says, hostnamelookups will add latency to every request and the default setting is off. But, when I first comment it out instead of off, the cpu usage increases. Below is the result. Why would
there be any difference between it being "off" or just commented out?
We have been using "restart" for Apache after the config changes. Then we tried "stop" "start", although we did them quickly. To get a good comparison of the Directives we are testing should stop and start Apache each time and waiting some suggested amount
of time? Like maybe our TimeOut setting or longer?
The values in the table are for the cumulative cpu and ram resources used by all httpd processes at that moment.
I guess we expected a wider variance in the values, and wondering if our testing process is the reason there isn't.
Thanks
|