Reading through the docs on clustering and I'm unclear on the mechanism and
requirements... Which of the following is the correct implementation, or are
there other options?

1. Clustering the UserApp requires a separate device (a load balancer) to
"front" an IP address and distribute requests to the UserApplication
according to some algorithm (round robin, load balanced, fail over only,

2. One of UserApp cluster nodes is configured as the "dispatcher" and takes
all requests, parceling them out to the other nodes in the cluster. In
effect, the load balancer is built into the JBOSS implementation. I've seen
references on the JBOSS forums to various built-in load balancers. (Note,
this seems to defeat the purpose of clustering if the hardware or IP
connectivity to this one device goes down... then no other cluster nodes can
service requests.)

Thanks in advance for any clarification you can offer.