Query optimization in star computer networks

Abstract

Query processing is investigated for relational databases distributed over several computers organized in a star network. Minimal response-time processing strategies are presented for queries involving the select, project, and join commands. These strategies depend on system parameters such as communication costs and different machine processing speeds; database parameters such as relation cardinality and file size; and query parameters such as estimates of the size and number of tuples in the result relation. The optimal strategies specify relation preparation processes, the shipping strategy, serial or parallel processing, and, where applicable, the site of join filtering and merging. Strategies for optimizing select and join queries have been implemented and tested.