One more thought. You are varying quite a few things as you look for a bottleneck.
Perhaps you could work in a more simple iterative manner. Start with a simple app and if possible use our load balancer. Confirm that this configuration works ok. Then start changing "variables" from this successful point - for example
a) use the same simple app but now use your load balancer. If you see a difference, then you have a strong clue that the load balancer is key.
b) then perhaps try more complex apps - see at what point things start to go wrong.
This may relate back to my earlier response about the size of data being replicated.
In general it is easier and you can learn more, more quickly by starting from a working system and then adding changes until you find the key change. Then it becomes a bit more possible to get to root cause(s).
[Message sent by forum member 'lwhite' (lwhite)]
http://forums.java.net/jive/thread.jspa?messageID=231925