Distributed join processing using bipartite graphs

Distributed query processing algorithms usually perform data reduction by using a semijoin program but the problem with these approaches is that they still require an explicit join of the reduced relations an the final phase. We introduce an efficient algorithm for join processing in distributed database systems that makes use of bipartite graphs in order to reduce data communication costs and local processing costs. The bipartite graphs represent the tuples that can be joined in two relations taking into account also the reduction state of the relations. This algorithm fully reduces the relations at each site. We then present a partitioning algorithm for response time optimization that takes into account the system configuration, i.e., the additional resources available. We also report on the results of a set of experiments that show that our algorithms outperform a number of the recently proposed methods for total processing time and response time minimization.