DYNAMIC LOAD SHARING IN DISTRIBUTED DATABASE SYSTEMS.

Abstract

The authors investigate dynamic load-sharing strategies for locally distributed systems in which the database is partitioned and distributed among multiple transaction processing systems and a common front-end processor is used for transaction routing. In this environment, if a transaction issues a database request referencing a nonlocal database partition, the request has to be shipped to the system owning the referenced partition for processing. A total of four different dynamic strategies have been studied. Their performances are compared with that of the optimal static strategy. A dynamic load-sharing strategy which takes into account previous transaction routing decisions is found to provide a substantial improvement over the optimal static strategy. The robustness of this strategy is further studied through sensitivity analysis over a wide range of conditions, including transaction load, communication overhead, and database reference distribution. 20 refs.