SharesSkew: An Algorithm to Handle Skew for Joins in MapReduce

Abstract

In this paper, we investigate the problem of computing a multiway join in one round of MapReduce when the data may be skewed. We optimize on communication cost, i.e., the amount of data that is transferred from the mappers to the reducers. We identify join attributes values that appear very frequently, Heavy Hitters (HH). We distribute HH valued records to reducers avoiding skew by using an adaptation of the Shares~\cite{AfUl} algorithm to achieve minimum communication cost. Our algorithm is implemented for experimentation and is offered as open source software. Furthermore, we investigate a class of multiway joins for which a simpler variant of the algorithm can handle skew. We offer closed forms for computing the parameters of the algorithm for chain and symmetric joins.