摘要： 为提高数据层的任务处理性能，提出一种可配置任务生成方法。通过分析数据库运行日志和跟踪信息，按照负载时空特性生成数据层任务处理脚本，并获取对应的资源消耗模式和关键任务处理性能。测试结果表明，该方法可以模拟实际应用系统的执行情况，由此获得服务器资源占用率，该资源占用率与实际资源占用率之间的误差可线性消除，为数据库服务器容量规划、存储结构设计、关键功能设计、性能优化，以及云计算平台资源实现和计算能力按需优化配置提供依据。

关键词: 负载特性, 资源消耗, 承载能力, 容量规划, 资源整合, 数据与计算双重密集型任务

Abstract: In order to improve the task processing performance of data layer, this paper presents a method to generate the tasks. The scripts of tasks are generated in accordance with the load characteristics or special requirements. By analyzing database running log and trace information, it gets the data processing tasks corresponding resource consumption patterns and processing performance. Test results show that this method can simulate the practical application of the system implementation, thus accesses to server resources occupancy linear eliminate the error between the acquired resources occupancy rate and the actual resources occupancy. This data can provide the basis for cloud computing platform to achieve optimal allocation of resources and computing power in accordance with the demand, and provide a basis for capacity planning of the database server, the storage structure design, the key features of the design and performance optimization.

Key words: load characteristic, resource consumption, carrying capacity, capacity planning, resource integration, data and computing double intensive task

中图分类号: 

• TP311