INCORPORATING PROCESSOR COSTS IN OPTIMIZING THE DISTRIBUTED EXECUTION OF JOIN QUERIES

The processing of a join query in a distributed environment exacts the usage of both the network and its computational facilities. A formulation that accounts for both, and felicitously constructed as an integer linear program, is proposed. Information disseminated among the sites of a distributed system is to be amalgamated and presented to a user, in response to his request. From all possible strategies by which this might be achieved, one necessitating the smallest usage of system resources is to be chosen. The data transferal resources of the network are usually presumed to be of greatest significance, and therefore, an optimal strategy is most often defined to be one which minimizes the total transmission cost. One model conforming to this philosophy, appearing in [1), expediently takes the form of a linear integer program, and so forms the basis for further refinement. In the omission of processing costs, the various processor elements of the network are treated homologously; dissimilarities in processing ability are also ignored. By analyzing the nature of join computations at a single processor, the minimal transmission cost model can be hybridized to incorporate the cost of such computations and differences in processing power.