Hello all my first post here , so please be gentle... :) I am a newbie to memcache....
I am the sys admin for a company that does online movie streaming. Our current setup is 6 Linux Centos web servers running apache 2.x These webservers sit behind HA Proxy and reverse squid proxy. We recently acquired some additional traffic to our site... but at peek times we seem to hit a plateau of sorts. We have 3 instances x 2 gigs of memcache on each of these web servers, one for sessions, two for mysql, so thats 18 instances in all, they all communicate with each other over the same NIC as all other traffic. Around 4pm to 6pm we get more traffic from Europe, during this time our servers go up to a load of 100 or more... Our hardware are dual quad core xeons with 12 gigs of ddr3 RAM and hardware raid 5. Our hosting company said we did not need more ram as its DDR3 and its used more efficiently than non DDR3 RAM, during our high loads our RAM does not go to low. Our connection limit is set to 4100 for all memcache instances. But we seem to reach a point where we cannot go more than 3,560 ?? Now can there be something else that might limit this? Are there other options i should be configuring in memached? SHould i be using a seperate NIC just for memcached use? Apache servers are set up like this... <IfModule prefork.c> StartServers 75 MinSpareServers 50 MaxSpareServers 175 ServerLimit 600 MaxClients 600 MaxRequestsPerChild 15000 </IfModule> Any suggestions? Thanks..
