Modeling Network Latency and Parallel Processing in Distributed Database Design
研究了在分布式数据库设计中如何通过数据分配和查询处理策略的联合优化来利用并行处理，减少网络延迟和本地处理对响应时间的影响，并采用遗传算法同时优化两者。
ABSTRACT The design of responsive distributed database systems is a key concern for information systems managers. In high bandwidth networks latency and local processing are the most significant factors in query and update response time. Parallel processing can be used to minimize their effects, particularly if it is considered at design time. It is the judicious replication and placement of data within a network that enable parallelism to be effectively used. However, latency and parallel processing have largely been ignored in previous distributed database design approaches. We present a comprehensive approach to distributed database design that develops efficient combinations of data allocation and query processing strategies that take full advantage of parallelism. We use a genetic algorithm to enable the simultaneous optimization of data allocation and query processing strategies. We demonstrate that ignoring the effects of latency and parallelism at design time can result in the selection of unresponsive distributed database designs.