提出一种基于服务时间标记的性能异常侦测方法,标记相对稳定的请求服务时间,并通过分析该时间变化来定位性能瓶颈以及分析可能的原因。该方法已经实现在一个典型的多层web应用系统的支撑平台中,TPC.W测试基准的测试结果验证了其有效性。
This paper presented a method based on service time signature for performance anomaly detection. Utilizing the relatively stable characteristics of request service time, this method helped to locate performance bottlenecks and to provide possible reasons by analyzing the variety of service time signature. And Implemented this method in a typical supporting platform of multi-tier Web applications. The results of TPC-W benchmark that the method is efficient.