Once the wiring is in place, the rest is easy. Install nodes/desktops whereever you need them, hook them into the network (possibly using cables you made yourself out of the leftover wire and a box full of RJ45 crimp-on plugs), install linux including PVM and MPI and parallel tools and compilers, and have at it.
What can you do with it once you've built it? Well, for one thing, a home beowulf usually has a reason to exist as components alone. Each machine usually ``belongs'' somewhere and does several things. What you do with it along the lines of parallel computation depends on what you do and what you want to get out of it. I use mine for code development, as a testbed for parallel applications and management tools, and to a lesser extent as a production unit in my research. I don't need a lot of nodes or fast nodes or matched nodes for these purposes, but working at home frees my nodes at Duke for production and keeps buggy code off of important machines until it is pretty well debugged.