You can cluster almost anything--EJBs, JNDI, RMI--in this application. Most of the application’s services are EJBs, so by default they are almost entirely clusterable. After initiating a client/server connection, each client specifies its target partition by opening connections with the cluster, not with the individual server--in other words, by specifying the partition as the server host. This is so the client communicates with the cluster and not with a single server. If a client specified a server by its actual name or IP address, the client would connect only through that server and could not use the load-balancing capabilities of the cluster/partition.