I’ve been testing the preload feature for several weeks on a blog with over 20,000 posts. The OS is RedHat ES 5, running WordPress 2.9.2, and memcached to cache database transactions.
As the preload filled with posts the server load remained about the same, the memcached hits decreased by about 66%, and the RAM used increased significantly as the OS cached frequently used files in the preload cache.
The RAM usage wasn’t exactly expected, but it can be detrimental on a server with a lot of posts, but not a lot of RAM. The server I tested this on has a lot of RAM so it wasn’t a problem.
What made the biggest difference on server load for this blog was using memcached for database transactions, which is running on the same server. I also use APC on the same server to cache PHP scripts, which also helps decrease the server load.
The RAM usage by the preload cache adding files and directories on the disk decreases the amount of RAM available for heavy traffic loads.
I also noted that Google crawled less than half of my posts after less than a week of using a fully loaded preload cache. This negatively affected the amount of traffic to this blog.
I tested this on a highly trafficked blog to get the best results.
For this blog it is better not to use the preload, because it needs the RAM for heavy traffic spikes, Google sends less traffic, and Google crawls the site less perhaps because it believes the files are static so they haven’t changed.
Thank you for leaving us the option not to use preload.