8.2 WebSphere clustering on micropartitions

 

+

Search Tips   |   Advanced Search

 

IBM WAS Network Deployment V6 workload management optimizes the distribution of incoming requests between application servers that are able to handle a client request. WebSphere workload management is based on application server clusters containing multiple cluster members. An application deployed to a cluster runs on all cluster members concurrently. The workload is distributed based on weights that are assigned to each cluster member. Thus, more powerful machines receive more requests than smaller systems.

The simplest way to increase the throughput of a well-tuned and efficient WebSphere application is to give it more resources. In the traditional WebSphere world, this was accomplished in the following three ways.


Put the application onto a bigger machine

Instead of a dual CPU, this involves moving to a four-way box and instead of using Gb of memory, giving it 2 Gb of memory. Depending on the application, this may or may not be the right approach.


Create a vertical cluster

This involves having another instance of the application running on the same machine. Applications that require large amounts of memory may be limited in the amount of Java heap space they can use; in these instances, it is more efficient to have a second Java virtual machine to provide the extra memory.


Create a horizontal cluster

This involves having another machine running another instance of the application.

The same holds true in the world of logical partitions and virtualization, except that it is much easier to implement.

To put the application onto a bigger machine, all that is needed is to assign more CPU or memory to the LPAR.

To create a horizontal cluster, you simply need to create another LPAR and assign it some resources. With the clever use of scripting, you can automate much of this work and have extra performance when needed, with little effort.

LPARs simply the addition of new resources to a WebSphere cluster. In our scenario, under Dynamic testing, we demonstrate the effect that dynamically adding resources has on an application under significant load. This is significantly easier than configuring new machines into a cell or migrating to bigger machines, and has the added advantage that the overall environment is easier to manage.


Next