Tree structured multiple processor join methods

Abstract

This paper summarizes the execution cost of join operations performed by parallel executing processors. Four different parallel join algorithms are proposed for execution on multiple processing nodes interconnected on a tree structured communication network. Secondary storage is accessed in parallel by leaf nodes. An average execution cost analysis is presented for the multiple processor join methods. For a reasonable ratio of result to operand cardinality, joins which use hashing for semi-join of operands at the leaf node secondary storage interface are predicted to perform better than nested-loop and sort-merge joins. Both node memory capacity and join result cardinality have a large influence on the relative performance of the join methods. This analysis method forms a basis for selecting among alternative processing methods for statically linked, multiple processor computer architectures.