I had been trying to puzzle out if using persistence on my small load balancer would permit better usage of apache created processes if combined with persistent Apache designs.
My thinking is the fact that without persistence around the load balancer, a person could finish on the servers, while a clear case of Apache created would take a while to die being idle. Basically routed that same connection right through to exactly the same Apache process with persistent, then there's little wasted cpu on idle connections in the load balancer switching to a different server.
Or shall we be held seeing this wrong?
(this is a piece of content explaining the main difference between persistent and persistence http://devcentral.f5.com/weblogs/macvittie/archive/2008/07/11/3443.aspx)
Yes, you can enable sticky periods around the load balancer (IP-based) to ensure that the persistent connections aren't produced on multiple apache instances.
Also, if there is a firewall involving the Apache servers as well as your application-tier, you would like to match the firewall's timeout in your connection.