Deductive Query Processing With an Objective-oriented Semantic Network in a Massively Parallel Environment

Abstract

Most research related to parallel query processing has concentrated on how to properly partition and schedule operation-by-operation and tuple-by-tuple query processing jobs to available processors. As a result, because these operations should perform complex query optimization, tremendous overhead can be involved, especially in a massively parallel system with thousands of processors. Furthermore, there exist unnecessary dependencies among operations allocated in different processors, and a large amount of intermediate data must be exchanged among processors. This article proposes an effective deductive query processing method in a massively parallel system. For this, the facts and deductive rules of a deductive database are partitioned into fine-grain semantic elements based on the concepts of an object-oriented model. These semantic elements are used to construct an object-oriented semantic network (OOSN). Because all facts and deductive rules are mapped to the OOSN statically, a query can be evaluated effectively in a distributed manner without any complex query optimization.