An object-based query evaluation scheme for deductive databases in massively parallel computing environment

Abstract

An algorithm is introduced to evaluate queries in a massively parallel computing environment under an object-based representation of a database. By distributing facts into connected networks of processing elements, it is shown that parallel execution of a query can be achieved in a cooperative way, so that the conventional tuple-by-tuple evaluation can be avoided. Furthermore, it is shown that dependency imposed by shared variables is not a major problem in achieving AND parallelism any longer. Although only a single query is considered, the evaluation scheme can be extended to multiple queries with multiple query networks and with replication of counters in the fact network.<>