Processing theta-joins using MapReduce

12 years 7 months ago

Download www.ccs.neu.edu

Joins are essential for many data analysis tasks, but are not supported directly by the MapReduce paradigm. While there has been progress on equi-joins, implementation of join algorithms in MapReduce in general is not suﬃciently understood. We study the problem of how to map arbitrary join conditions to Map and Reduce functions, i.e., a parallel infrastructure that controls data ﬂow based on key-equality only. Our proposed join model simpliﬁes creation of and reasoning about joins in MapReduce. Using this model, we derive a surprisingly simple randomized algorithm, called 1Bucket-Theta, for implementing arbitrary joins (theta-joins) in a single MapReduce job. This algorithm only requires minimal statistics (input cardinality) and we provide evidence that for a variety of join problems, it is either close to optimal or the best possible option. For some of the problems where 1-Bucket-Theta is not the best choice, we show how to achieve better performance by exploiting additional ...

Alper Okcan, Mirek Riedewald

Real-time Traffic