Parallel Processing of Recursive Queries in Distributed Architectures

This paper presents a parallel algorithm for recursive query processing and shows how it can be efficiently implemented in a local computer network. The algorithm relies on an interpretive approach where recursive rule processing and data retrieval are merged in a top-down computation. It employs “sideways information passing” to restrict to relevant facts the information extracted from the relational database. Evaluation is divided into a compilation phase and a dynamic phase. The compilation phase statically constructs a derivation tree that expresses the decomposition of a query into subqueries and the “sideways information passing” strategy. In the dynamic phase, cooperative processes are associated with subsets of “equivalent” nodes of the derivation tree. They communicate by message passing without sharing memory. Some optimizations are discussed for a practical parallel implementation. Gains in efficiency with respect to classical sequential algorithms are also disc...

• Type: Conference
• Year: 1989
• Where: VLDB
• Authors: Guy Hulin
• Keywords: Compilation Phase, Database, Dynamic Phase, Sideways Information, VLDB 1989