I am trying to deploy a Wordpress Appliance (wordpress-11.2-lucid-x86) on a medium AWS server (c1.medium, 5 ECU, 1.7G RAM) and am having trouble running under load. By default, the appliance is using the prefork apache install, with each apache2 process using 60-63MB of memory. I've already disabled every apache module that isn't needed by wordpress and brought this memory usage down to 55-60MB. By default the MaxClient setting is 150 which would be over 8GB memory usage. I had to increase the MaxClient setting (at 250 right now) to get page load times to be under 10 seconds. But at either of these, I'm well beyond the 1.72GB ram that the medium AWS server has to work with. I'd have to bring MaxClients down to about 30, which is unacceptable for a production server. After running for a few hours, all access to the server (webmin, shell in a box, ssh, apache server) was unresponsive and I had to force a reboot from the TurnKey Linux Hub. I suspect apache finally used all the RAM and swap and brought everything to a grinding halt.
Has anyone gotten a Wordpress Appliance to run under moderatly heavy load? Is the apacher worker mpm version better than prefork for this? Are there Wordpress/PHP/Apache settings that are key to getting acceptable performance?