The owning server instance makes the timeout callbacks so in the typical case the timeouts will happen in the same JVM in which createTimeout is called. There's no special attempt at the timer service level to load balance the timers.
However, the load is typically balanced anyway since in the cluster case the client requests themselves, whether entering via HTTP, JMS, or the EJB Remote view, are usually configured to be spread across the server instances.
[Message sent by forum member 'ksak' (ksak)]
http://forums.java.net/jive/thread.jspa?messageID=321494