Forum Moderators: phranque
I have been running an Apache2 site on my dedicated hosting for a few years but it's showing its age. My host now recommends that now that I have high traffic I consider a redundant load balancer (hardware).
This means I would have to maintain 2-3 machines with identical web content. Isn't this a pain? To ftp to three machines everytime? How do people do it -- with scripts?
Secondly, the load balancer is priced by connections per second. I'm thinking 200 per second or 500 per second. But my Apache2 is currently set up as:
KeepAlive On
KeepAliveTimeout 2
MaxKeepAliveRequests 300
MinSpareServers 5
MaxSpareServers 12
StartServers 5
MaxClients 250
MaxRequestsPerChild 600
And it works fine. Isn't this a lot more than the load balancer type "200 connections per second"? Or are these two metrics even equivalent? How should I measure how many per-second connections my Apache gets?
Thanks
Do you know if your bottleneck is Apache or is it the database? Often times Apache is waiting on the database, and quite capable of handling more traffic if only the db could keep up.
If this is your situatio, before ramping up to multiples load balanced servers I would look into integrating a caching solution like memcached. It helped my site handle higher loads while sticking with one server. You can also scale out your cache over multiple servers if needed as well. Though so far I'm still on one server. We handle around 5 million pageviews a month. Not huge traffic, but it definitely would push our server to it's limits before we implemented memcached.
Google it up to learn more. It is free and used by many big name sites like Youtube, Facebook, etc.