Query Processing in Distributed Database Systems

Abstract

Query processing on a distributed database system requires transmission of data between computers on a communication network. Minimizing the amount of data transmission is one of the fundamental principles to reduce the query processing cost and to prevent network congestion. The semijoin operation is important in formulating query processing strategies. Semijoin preprocessing strategy provides enough opportunities to significantly reduce the amount of data required to be transmitted on the network by first reducing the cardinalities of a distributed relations using semijoins and then transmitting the resultant relations to the result node. Algorithm LIGHT which is based on new heuristic for generating semijoin preprocessing strategies for queries has been developed in this work. Heuristic used in algorithm LIGHT can be thought as modification of heuristic used in algorithm SDD1. Simulation model has been constructed to evaluate the performance of existing query preprocessing algorithm SDD-1 and algorithm LIGHT. This model test random queries and results are presented and discussed. It is shown that algorithm LIGHT performs better than algorithm SDD-1 in general.