6.5 Clustering

Clustering OC4J instances allows Oracle Application Server to solve both hardware and software reliability and scalability problems. Oracle Application Server can group two or more OC4J instances into a cluster, whether or not they reside on the same host. Clustered OC4J instances appear to the outside world as one application server, providing high reliability and almost unlimited scalability.

Various load-balancing mechanisms are available to direct incoming requests between clustered OC4J instances. If one clustered instance fails, another picks up where it left off, using the appropriate load-balancing mechanism. Clustering OC4J instances across hosts solves both the hardware and software reliability problems. The bulk of what clustering is and how it is employed in Oracle Application Server was covered in Chapter 2, but a few relevant J2EE issues are discussed in this section.

OC4J provides stateful failover for clusters via HTTP session object and EJB state replication. This means that each user 's session, and each EJB's state, is replicated between OC4J instances.

The HTTP session objects in a web (servlet) container are selectively replicated across all OC4J instances that are part of the same cluster and processes that are part of the same island. An OC4J island is a way to group one or more OC4J JVM processes. Rather than having to replicate HTTP session objects among all OC4J processes (which can be very resource- intensive ), islands allow selective replication. This fine-grained control prevents replication within a cluster from degrading OC4J's performance.

EJB state replication is handled differently. Because of the nature of EJBs ”they are all independent components ”EJB state is replicated between all OC4J instances in a cluster. There is no fine-grained replication control mechanism.