An Improved Hash-based Join Algorithm in the Presence of Double Skew on a Hypercube Computer

This paper presents an improved parallel hash-based join algorithm on a hypercube computer in the presence of double skew. We describe a load balancing technique to evenly distribute both join relations across all processors in order to deal with double skew eeectively. Moreover, we propose a permutation join method which reduces main memory requirement for the local join operation in the previous method presented in 8] and speeds up the local join operation in each processor by using a two-index join method. Our algorithm partitions the relations into a number of buckets, evenly distributes each bucket to every processor by using the load balancing technique, and then joins each pair of matching buckets simultaneously by using the permutation join method. The performance analysis of the algorithm shows that our algorithm is more eecient than the previous algorithm 8] in the presence of double skew.