Performance analysis of distributed database processing in bus local networks

As the number of local area network (LAN) installations rapidly increases, it is essential to analyze the end-to-end performance of LAN-supported information processing. Due to its high bandwidth and limited span, a LAN creates a loosely-coupled parallel processing environment for information processing in a local area. The performance level of such information processing cannot be implied merely from network performance, however. This paper analyzes the database (DB) transaction response time of bus-based distributed database systems using detailed queueing models of bus transmission and local database subsystems. All the analytic results have been verified using simulation models, and the impact of transaction intensity levels and allocation policies on the response time performance and the accuracy of analytic results have been examined. The results show that the analytic performance model closely evaluates the transaction response time over a range of transaction intensity levels. In addition, the results reveal that in a typical data processing oriented system, the bottlenecks lie in disk access systems in local areas. As a result, high replication policies for DB allocation design result in improved response time performance in bus networks.