MRTree: Functional Testing based on MapReduce´s execution behaviour

Abstract

MapReduce is a paradigm that allows parallel processing of large amounts of data. MapReduce programs combined with their underlying run-time framework have distinctive features that are prone to include unexpected behaviors not present in other types of programs. This paper describes an approach to functional testing of MapReduce programs based on a hierarchical classification of a number of potential faults that may occur in MapReduce programs over Hadoop. This classification, called MRTree, is then used to derive test cases able to detect the faults represented in MRTree and illustrated with some examples