A cyclic multi-relation semijoin operation for query optimization in distributed databases

In this paper we have introduced the concept of multirelation semijoin and we have shown that it is possible to get substantial reduction in the cost of data communication if a restricted form of this operation called cyclic multi-relation semijoin is used for query optimization in distributed databases. We have proposed a heuristic to identify situations where the cyclic multi-relation semijoin operation is likely to be useful. To study the applicability of this operation and to determine how much additional benefit we may expect, we have augmented a well known semijoin based algorithm with this operation. We have used simulation studies with a large number of queries and our experiments indicate that depending on the characteristics of the database, improvements ranging from 30% to 80% is possible