Parallel evaluation of a deductive query with a breadth-first search strategy

The query evaluation in a deductive database system requires not only a number of deductive laws to be processed but also a large volume of facts to be matched. To achieve more parallelism embedded in the logical representation of deductive databases, this paper proposes a parallel execution method that expands an AND-OR query tree with a breadth-first strategy. In order to cope with the recursive invocation of a query, the AND/OR tree is expanded in an incremental way. As a result, the two major operations, fact-matching and law-expanding, of the query evaluation can be performed separately. Compared with other approaches to parallelize the query evaluation process, the proposed method eliminates the producer-consumer dependency among shared variables and provides a convenient way to collect all alternative answers by the late binding of the variables in a query.