I use Hadoop natively on Windows as a virtual 2-node cluster running on one machine. It runs inside Cygwin (so no VM). Works well to try Hadoop out and I still use it to test new code in small before putting it on the cluster. You basically get every bit of functionality as with a full cluster. Getting it to work can be a bit tricky though.
I used the following short guide: Stanford Hadoop for Windows guide
Which worked fine for me. Very important is that you use 0.20.0! Higher version do not run under Cygwin. I think it is best to leave the number of default nodes to 2. This way you can test if splitting the work across multiple nodes works, but more simultaneous nodes can give you memory problems.
