Should a node not be able to fulfill its duties for any reason, another mode could step in and take care of the work and data. This is called failover and is exemplified in the following diagram (true or silent failover is completely transparent to the client):

Failover: Should a node in a cluster leave the cluster for some reason a client call can be handled by another node in the cluster

Should the failing node come back to work again after a period of time, work and data can be transferred back again by a failback procedure. This transfer often has some restrictions, for example, it can only happen during the next call from the client.

Data chosen for replication in Java EE solutions most often describes the session state. A session is a set of data that belongs to a particular communication flow. It can be persisted but more commonly resides in faster and more volatile memory (RAM). A common example is the HttpSession, where each instance with content...