Experiences with distributed query processing

Different implementations of an experimental distributed query processing system which is constructed on top of existing database management systems (DBMSs) are presented. A performance evaluation is carried out. It is shown that the response times of query execution are affected significantly by the degree of site autonomy, interferences among processes, interface with the local DBMSs, and communications facilities. Too much autonomy for data replication increases response times. However lack of processing autonomy also increases response times. A goal is to identify the type of autonomy that facilitates query processing. It is shown that pipelining of the actions for data replication, which include local reduction, data transfer, and union fragments, improves performance whereas pipelining between join and data replication causes deterioration in performance because of competition of resources. However, the simulation results indicate that, in a multiprocessor environment, pipelining always improves performance provided that processes which compete for the same resource are assigned to different processors