This week's book giveaway is in the OO, Patterns, UML and Refactoring forum. We're giving away four copies of Refactoring for Software Design Smells: Managing Technical Debt and have Girish Suryanarayana, Ganesh Samarthyam & Tushar Sharma on-line! See this thread for details.
Well, first, there is the architecture aspect. You have to install WebSphere on multiple machines, and make those machines part of a common administrative domain, or cell. Essentially, you have a bunch of WebSPhere nodes that host applications, and those nodes are managed by one central deployment manager.
However, that's not clustering - that's just architecture.
Then, you create a cluster accorss your WebSphere nodes. Essentially, you put a JVM on each WebSphere node, and tell the deployment manager that all of those JVMs will host the same applications.
THEN, you deploy as many EAR or WAR files to the cluster, and WebSphere magically deploys the application to each JVM. At this point, each JVM can handle the requests that come in from the web client. The HTTP server gets a plugin-cfg.xml file that sprays requests randomly/round robin to each of the JVMs running on the WebSphere nodes.
You should note that at this point, you are not workload managing HttpSessions or stateful session EJBs (SFSBs). This is some extra steps that must be done through the admin console, and can be accomplished through database persistence, or in memory replication.
Anyways, it's a big question.
I have an article on my website describing WebSphere Clustering and deployment options:
I think those will answer a heck of alot of questions for you regarding WebSphere. The multimedia tutorials use WebSphere 5 right now, but there is very little difference in thi aspect between WebSphere 5 and WebSphere 6.1. Who knows what will happen with WebSphere 7?