Planning the Architecture of a Cluster

The configuration of SAP J2EE Engine 6.20 cluster must be planned considering two basic factors:

Deployed and running applications – the number of server nodes must change relative to the number of applications. The leading criterion is the total consumption of calculating and other system resources for all applications to be accessed by clients.
Number of client requests – the number of dispatcher nodes must increase proportionally if there is significant growth in the number of client requests.

Cluster Scenarios

There are several scenarios for SAP J2EE Engine 6.20 cluster configuration because each configuration must include at least one dispatcher; otherwise, no cluster can exist:

Single server, single dispatcher – this is the initial configuration of SAP J2EE Engine 6.20 cluster. This solution is appropriate for a small number of similar applications, which do not increase the load factor and requests processing of the single server node.
Single dispatcher, n-server configuration – this is probably the optimal solution for a great number of diverse applications. In this configuration, all client requests are sent to the dispatcher, which then redirects them to the least-busy server for processing. When the number of requests is increased significantly, new server nodes can be added to achieve greater scalability and to reduce response time. This configuration enables optimal load balancing because the only function of a dispatcher is to reroute client requests.
m-dispatcher, n-server configuration – this is the cluster scenario in which the greatest number of requests can be handled. This configuration is appropriate when each dispatcher node is to receive client requests for a particular type of application. The optimal reduction of response time to client requests can be achieved because each dispatcher maintains connections to each server in the cluster. In this way, the load factor of the particular module on each server can be balanced easily.