Query processing in distributed databases with nondisjoint data

Abstract

In a distributed relational database, relations are divided into disjoint fragments. These fragments are allocated to different sites in the database using some allocation scheme to improve the data retrieval time. Allocation schemes that are not constrained by other features of the DBMS will make the allocation easier to implement and provide desired performance (retrieval time). However, at present, the allocation schemes are constrained by the assumptions made by the existing query processing schemes. Most of the existing query processing schemes assume a restricted form of fragment allocation. Some of these assume that a fragment is allocated to only one site, while others assume that the sets of fragments allocated to two different sites are either disjoint or the same. This paper emphasizes the importance of nondisjoint data among sites in a distributed database environment and presents query processing framework for such an allocation. Using the framework, a number of query processing schemes can be implemented. This paper also presents a heuristic query processing scheme using this framework. The heuristics presented here attempts to use the redundant data to eliminate the expensive join, I/O, and communication cost.