通用海量数据库性能测试系统的设计与实现

General performance measures for massive databases

为客观反映数据库实际运行性能,设计一种可配置数据库性能测试标准,它能够针对具体应用数据库系统,建立相应的控制模型、数据模型和事务处理模型。控制模型描述数据库的运行模式,数据模型描述用户访问数据库的数据集合,事务处理模型描述用户访问数据库的行为模式。控制模型通过事务处理模型和数据模型组装动态SQL(standardquery language)语句,然后产生多个并发进程模拟大量实际用户访问数据库,以测试数据库的运行状况和性能。分析与实际测试结果表明,该系统能够根据实际应用系统配置数据库性能测试标准,较之目前国际上通用的固定模型测试标准如TPC-C等,更能够客观反映数据库实际运行性能。

A configurable database performance benchmark was developed to evaluate database performance. The benchmark system provides a control model, a data model and a transaction processing model for specific application systems. The control model provides the database operating mode. The data model provides a data aggregate when a user accesses the database. The transaction processing model provides the behavior mode for database users. The dynamic SQL （standard query language） statements which simulate the actual users are assembled by the control model via the transaction processing model and the data model. Then, a large number of visits to the databases are simulated to analyze the performance of the databases. Analysis and testing results indicate that the system can configure the database performance benchmarks to more objectively reflect the actual database performance than commonly used fixed database benchmarks such as TPC-C.