There is one minor requirement before thinking about allocation of hardware resources: defining and selecting the hardware used for your cluster. In this section, we'll discuss a physical cluster and move on to Amazon EMR in the next.

Any specific hardware advice will be out of date the moment it is written. We advise perusing the websites of the various Hadoop distribution vendors as they regularly write new articles on the currently recommended configurations.

Instead of telling you how many cores or GB of memory you need, we'll look at hardware selection at a slightly higher level. The first thing to realize is that the hosts running your Hadoop cluster will most likely look very different from the rest of your enterprise. Hadoop is optimized for low(er) cost hardware, so instead of seeing a small number of very large servers, expect to see a larger number of machines with fewer enterprise reliability features. But don't think that Hadoop will run great on any...