Two roads diverged in a yellow wood. I took the one less traveled by, and that has made all the difference.
In this chapter we look at different ways of virtualizing the worker nodes in a Hadoop cluster—those nodes that contain the NodeManager, Application Master, Containers, and DataNode processes as their Hadoop roles. The standard native configuration that you might have seen up to now has been to combine compute processing (NodeManager, Application Master, Container) and storage (DataNode) on each Hadoop worker node. We will initially use that approach here when we virtualize the worker nodes and then show how you can go beyond it for flexibility reasons.
The design approaches ...