The following figure depicts the client/server with isolated replicators topology. In this topology, the replicators are separated from the backup servers and are on their own servers.
There is a tier of applications servers that host web applications using HTTP Session, and these sessions are replicated out as they are created and updated. There is a second tier of servers without a web application installed, where the session manager receives updates from the replication clients. The replicators facilitating the transfer of data reside with the replication servers
Benefits of this topology include:
Although this topology is the most expensive from a Java Virtual Machine (JVM) perspective, and requires the most configuration through the administrative console, it provides the greatest degree of isolation from failure of a component, the most flexibility in starting/stopping the parts, and the most flexibility in mixing the right number of clients, servers, and replicators. Since both servers and clients need to connect to the replicators, starting the replicators removes the need for connection retries. Similarly, starting the clients after the servers, avoid timing windows where the clients send data before the servers are ready to receive them.