Properties of Semijoin Sequences

The problem of finding optimum semijoin sequence for an arbitrary query under linear cost function for the transmission cost is NP-hard. Hence heuristic algorithms with desirable properties are explored. In this paper four properties of semijoin programs for distributed query processing are identified. The use of these properties in constructing semijoin sequence is justified. An existing algorithm is modified incorporating these properties. Empirical comparison with existing algorithms shows the superiority of the proposed algorithm.