Parallel decomposition of relational operations based on fragmented column indices

This paper is a continuation and development of the previous our work, where the decompositions of intersection and join operations for columnar indices on the basis of domain-interval fragmentation were proposed. This decomposition allows to organize the parallel execution of relational operations on the distributed columnar indices without massive data exchange between the processor nodes. In this paper the decompositions of the projection, selection, duplicate elimination and union operations are considered. Also, a new kind of columnar indices called columnar hash-indices is introduced. The columnar hash-index can index multiple attributes of a relation. For distributed columnar hash-indices, the decompositions of intersection, union, and natural join operations are considered.