Where Do We Start?
There's a folk saying that goes something like this: "A map does you no good if you don't know where you are." The steps in a process are just directional signs on the road to building a cluster. There may be many routes from nowhere to a finished cluster, but taking a completely different path for every project may be wasteful.
The first task, if you're doing this for the first time, is to figure out a starting point. When confronted with a complex task, I generally revert to a "context diagram" of whatever I'm building. This technique is useful for uncovering some of the relationships between an object and its environment. (I guess this is just the old software SA/SD training drilled into me as a young programmer/analyst. See why I object to the "p word" so much?) Figure 1 shows a sample of a cluster context diagram.
Figure 1 Cluster context diagram example.
The context diagram shows the physical cluster interacting with its environment. Note that we may not be able to specify the parameters yet, although it's likely that some of them may be forced on us—space, heating, power, and other physical resources may be limited. We need to make a first-pass design and then improve it by checking it against requirements. This is where our old friend, the waterfall methodology, flows over a big precipice and hits the rocks below. This cluster-building process thing is by nature an iterative activity.