News for hackers. Stuff that matters.
Re:10 Million hits a day with Wordpress using a $15 server by sanswork on Friday March 30, 2012 @09:37AM (#3775844)
Try it again with just the last step(vanilla apache/vanilla wordpress/varnish) what are your results like?

Keep the original setup and have your test script perform some random searches. New figures?

(Everything is fast/non-server intensive when you're serving static data)

Re:10 Million hits a day with Wordpress using a $15 server by EwanToo on Friday March 30, 2012 @10:03AM (#3775946)
Just tried:

This rush generated 7,363 successful hits in 1.0 min and we transferred 56.39 MB of data in and out of your app. The average hit rate of 117/second translates to about 10,159,286 hits/day.

So you're right in that Varnish is the big improvement, but the CPU of the server seemed to be significantly higher with varnish alone than with varnish + APC.

Of course, one issue with these systems is there's no such thing as vanilla, the documentation goes from "Install" to "Read 100 pages of stuff to get a working configuration", with nothing in between...

Re:10 Million hits a day with Wordpress using a $15 server by dugmartin on Friday March 30, 2012 @09:41AM (#3775859)
I think the main difference between using Apache vs nginx as the backend server would be the amount of memory available for Varnish to use for caching. nginx has a much smaller memory footprint.
Re:10 Million hits a day with Wordpress using a $15 server by cd34 on Friday March 30, 2012 @10:15AM (#3776010)
apache-prefork has a larger footprint because it includes the php interpreter in each thread. If he's using php-fpm, he would use mpm-worker in which case apache and nginx threads have a similar memory footprint.