Configured Services - Clustering
ClusteRing While the concepts and the goal remain very similar, there can be significant differences in the way clustering is configured in WebLogic and JBoss. WebLogic provides a very static view of clustering, both in terms of cluster membership and configuration details. There are limited configuration options and clustering largely remains a black box. This leads to a level of simplicity that works well for some clients but can cost others with more exceptional requirements. The cluster itself is statically defined on the administration server with provided addresses of each member. In JBoss, a cluster is a very dynamic concept. It is defined by a cluster name and an address, which may be a multicast or unicast address. Any JBoss instance declaring an intention to join the cluster can do so by communication with that address and by providing the cluster name. This results in a very fluid cluster definition that works very well with dynamic provisioning scenarios. Furthermore, clustering in JBoss is built on top of JGroups, which itself is highly configurable and has configuration parameters that are exposed to cluster users. A wide variety of network topologies can be supported and the ability to fine-tune the cluster performance is virtually endless.