Map-reduce is a framework to process large data sets, splitting them into subsets, processing each subset on a different server and then blending results obtained on each.