Multi-join algorithm utilizing sub lattice of a minimal number of double indices

Abstract

In this paper, a novel multi-join algorithm is introduced. The novel algorithm is based on building a sublattice of a minimal number of double indices. A double index is a data structure that combines the indices of two relations and is built by a hashed-based join algorithm. The double index is divided into join buckets with the same hash function value. The algorithm then joins buckets with similar hash function to produce joined buckets. These steps are used to build the complete join index of the two relations. The time complexity required to build the join index of two categories is in the order of m log m where m is the size of each category. The time complexity to join all buckets is of the order of n log m. The join index is used to materialize the joined relation if required. Otherwise, it is used along with other double indices of other relations to build a sublattice to be used in multi-join operations with minimal I/O requirements. The sublattice of the double indices can be fitted into the main memory which further reduces time complexity of the multi-join algorithm. The main advantage of our technique is when used in intensive multi-join dominant computational environment.