Optimizing Distributed Joins with Bloom Filters

Distributed joins have gained importance in the past decade, mainly due to the increased number of available data sources on the Internet. In this work we extend Bloomjoin, the state of the art algorithm for distributed joins, so that it minimizes the network usage for the query execution based on database statistics. We present 4 extensions of the algorithm, and construct a query optimizer for selecting the best extension for each query. Our theoretical analysis and experimental evaluation shows significant network cost savings compared to the original Bloomjoin algorithm.