Managing the WAS cluster workload

Edit online

When clustering WebSphere® Application Servers that host Web containers, a plug-in workload management is automatically enabled among the cluster members.

Using the HTTP transport protocol, the routing of servlet requests occurs between:
  • The Web server plug-in (IHS plug-in for the SUT).
  • WAS cluster members.
Furthermore, there is the failover aspect of a WAS cluster environment:
  • If one server is unavailable or cannot perform the task, tasks can be reassigned to another cluster member.
  • Reassigning a task has the obvious advantage over running a standalone application server in that a standalone application server can become overloaded if too many requests are made.
Figure 1. Using an IHS to manage Content Navigator WAS cluster workloads

This graphic provides an overview of how an IBM HTTP Server (IHS) can be used for the management of Content Navigator WAS cluster workloads

In a WAS cluster, each member contains the same applications. Therefore, client requests can be distributed across cluster members according to the capacities of the different machines. For that purpose, weights can be assigned to each cluster member.

For the SUT, FNP8 CPE and ICN formed an application pair on an ECM node.
  • Each of the clustered applications ran in their own WAS cluster but on the same ECM node.
  • Workload management was only done only at ICN cluster level because ICN received client requests and was mapped at cluster level to the IHS Web server.
  • Due to the one-to-one relationship between the ICN and FNP8 CPE application on each ECM node, there was also an inherited workload management for FNP8 CPE.

The WAS cluster workload management is done by routing servlet requests to all available cluster members according to their weights.

For the SUT: The virtual machine capacities for the ECM nodes were equal. Therefore, each ECM node had the same cluster member weight assigned (the aim was to have equal workload distributions on all four ECM nodes).