Home | N-way peer to peer | WAS v5 to v6 migration
Single replica
The single replica configuration scales much better in terms of memory consumption. In the event of a failure of an application server, the request will be routed to the next clustered application server in the rotation. If the session information exists in the new server, it is used; if not, the session manager retrieves it from wherever it was backed up. As soon as the request hits the new server, its updated session information is sent to another server, so there is always a backup. (Unless the wrong two servers happen to fail at the same time.)
Client Server to Client Server topology is one where another application server is configured to store backup session data in the local memory space.
This topology reduces the overhead on application servers handling requests, but introduces a single point of failure (SPOF). It is useful to configure two independent application servers to store backup session information. By introducing a second server on a second machine, the single point of failure is eliminated.
A replication domain consists of servers and cluster members that have the capability to replicate information from one cluster member to any other cluster member.
Some changes in WebSphere Application Server v6 include cooperation between the Data Replication Service and the Workload Management subsystem to coordinate which cluster members serve as backups for other cluster members.
Ideally, session failover data and stateful session bean failover data should end up in the same place - the place that a failed-over session will arrive in the event it needs to be served by a cluster member other than the one that originated the session.
Another improvement is that the underlying mechanism has been rewritten using a proprietary transport to move data. This reduces overhead and improves overall system performance.