Formalizing MapReduce with CSP

Abstract

As a programming model, MapReduce is popularly and widely used in processing and generating large cluster of data sets distributed on large amount of machines. With its widespread use, its validity and other major properties need to be analyzed in a formal framework. In this paper, a formal model is presented using CSP method. We focus on the dominant parts of MapReduce and formalize them in detail. Through this formal model, the processing and function of each component can be clearly reflected. Moreover, we illustrate this formal model by an example computation. The result reflects the validity of MapReduce in some appropriate applications.