Query Processing in a Multidatabase System

Abstract

A multidatabase system provides a logically integrated view of existing, distributed, possibly inconsistent, databases that may be stored under heterogeneous database systems. Query processing in a multidatabase system consists of four tasks. First, queries over the user’s integrated view must be modified into equivalent queries over the local schemas of the existing databases. Second, the modified queries must be globally optimized to produce an efficient global execution plan, which specifies local processing steps at individual sites, communication steps iii which data is moved between sites, postprocessing steps in which the results of local processing steps are combined, and the order in which these steps must be performed. Third, the single-site queries to be executed in the local processing steps must be locally optimized to select efficient access paths. Finally, the optimized single-site queries must be translated into the data languages of the host database management systems. This paper address the first three of these problems.

Logical database integration is achieved primarily through the use of generalization, which is modelled algebraically as a sequence of outerjoins and aggregation operations. Conventional query modification and distributed query processing techniques are inadequate for such views. Previous research on distributed query processing techniques has focused on joins, ignoring selections and projections because they are easy and inexpensive to process in a conventional distributed database system. However, even selection and projection queries over views defined by outer-joins and aggregates can be expensive to process. We develop solutions to this problem. We also solve the problem of local query optimization for navigational database systems such as CODASYL. The main problem here is to compile queries in a high-level query-language into programs that use efficient navigational access paths for evaluating the queries.

This research was jointly supported by the Defense Advanced Research Projects Agency of the Department of Defense and the Naval Electronic Systems Command and was monitored by the Naval Electronic Systems Command under Contract No. N00039-82-C-0226. The views and conclusions contained in this document are those of the author and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the DARPA, the NAVELÉX, or the U.S. Government.