I'm having a different, but related, problem.
My site (www.palibandaily.com) is running the latest updates of everything. Has over 300 posts.
Normal traffic is about 2000-2500 unique visitors/day. It's been getting a bit sluggish, but tolerable.
Yesterday, an article hit Digg, and brought in about 30,000 unique visitors. The cpu was frequently throttled at the web host end (Bluehost), leading to downtime. I could not access the back end at all until it slowed down. (I'd had 45,000 hits in a day once, that was in July. Far less content at that time. No major disruption at that time.)
Their service folks informed me that there were several hundred thousand database queries in short periods. I figured that the modest traffic load couldn't cause that alone; his response:
Actually it does, since your using wordpress all settings, pages, links, images... all of those are mysql queries. wordpress is a very heavy mysql/php script that dynamically generates everything on the page. If you want your site to be efficient, you should make static HTML pages instead of load generating php processes that generate the HTML code for frequently accessed pages.
Your mysql Can't Connect errors are due to the amount of connections your site is attempting to make the database. For every page that's it viewed, refreshed, or clicked on, about a dozen or more mysql queries result to generate the page.
Each php process uses the cpu to generate the script, the more that do this, the more cpu is used and the more your account is throttled. More throttling occurs the longer it takes for the php process to execute. The longer the mysql is held open the more chance that the connection limit will be reached or the mysql will time out.
Therefore the amount of visitors to your site has an almost exponential growth factor to the performance limitations on your account.
I assume I'm not the only person who has been running into this problem, but either my forum-searching skills are inadequate, or there isn't much here.
Does anyone have any brilliant ideas to improve the efficiency?
Note: During the peak, the Web Optimizer plugin was running.