Query processing utilizing dependencies and horizontal decomposition

Since join operations are expensive, usually join scheduling is very important for query processing. In this paper we will discuss new procedures to handle cyclic queries utilizing dependencies and horizontal decompositions. There are three known procedures for cyclic query processing: (1) Relation merging, (2) Tuple-wise processing, (3) Attribute addition. As join operations are applied to relations which are processed by selection operations, the number of tuples is usually less than the original relation and thus there are situations in which temporary FDs are satisfied. Such FDs can be used to simplify the given query. To convert a given cyclic query into a tree, some relations must satisfy a set of FDs. This can be attained by horizontal decomposition. Tuple-wise processing and attribute addition are shown to be special cases of the FD-based procedure. We have also developed MVD-based procedures which are generalized from the FD-based procedure.