[wp-hackers] wp-cache2 performance (was apache 2.2.2 upgrade)

[Mark Jaquith]

On Jul 15, 2006, at 12:56 AM, Angsuman Chakraborty wrote:

> Exactly the same situation here. In fact I was as surprised as you  
> were with the results. We repeated the tests over several  
> configurations with same results.
> Then I understood. WordPress 2.0.3 already includes caching. It  
> just so happens that their caching mechanism alone is more  
> efficient than Wp-cache 2. I will share the results on Monday when  
> my developer who did the testing is back.

WP 2.0.3's built-in (but disabled) object cache is not a substitute  
for an output cache... they are different beasts.  Even with the  
object cache on and populated, you still need to make several DB hits  
for each WP request, and WP needs to be fully loaded for each  
request.  The object cache merely reduces the number of DB hits.  WP- 
Cache eliminates all DB hits for cached hits, and skips the full  
loading of WP.

> This is what I did:
> Selected a large amount of url's from the web site. Then took a  
> random subset and tested with one configuration, incrementally  
> increasing the number of threads. After completion I cleared the  
> cache and repeated for the other configuration. The results are  
> averaged. Several runs were made to rule out statistical errors.

This may not be a real world scenario.  After clearing the WP-Cache2  
cache, yes the server will work harder, as it dynamically serves  
pages and writes the output to disk.  But once the popular pages have  
been cached for the cookie-less user, the load will drop considerably  
because subsequent requests will be served from flat files, using  
very few CPU cycles.  It makes sense that a dynamic hit with WP- 
Cache2 would take more time than a dynamic hit without WP-Cache2...  
and if all your hits are unique, then WP-Cache2 isn't for you.   But  
the vast majority of sites serve lots of duplicate requests, and  
would greatly benefit from WP-Cache2.

--
Mark Jaquith
http://txfx.net/