Scaling Java JVM Application on AWS

I'm using a java .WAR application that I'm trying to make fault-tolerant and highly available. I'm currently running on Amazon AWS and using Auto-Scaling & Load Balancing. Problem is, Load-Balancing will only load-balance traffic, it won't know if the RAM is almost exhausted on a machine.

The problem occurs when we see a big surge of searches (which use the WAR), and eventually the JVM just crashes and tomcat has to be restarted. Anything I should be looking at to make my tomcat environment more available?