一种基于失效频度的微重启群判定方法

为减少递归重启过程中不必要的递归开销以实现应用系统的快速恢复，在微重启技术的基础上，提出一种微重启群的判定方法。该方法通过负载测试并在测试期间注入异常来获取组件的平均失效频度，以此分析组件间的失效关联程度，给出微重启群判定算法。研究结果表明，该方法可针对组件化分布式应用的故障进行重启，使系统平均恢复时间减少30%左右。

Decision Approach to Microreboot Group Based on Failure Frequency

To implement the fast recovery of application system by cutting unnecessary recursive overhead during the recursive restartability, a decision approach to microreboot group is proposed based on microreboot technology. By the aid of load testing and exception injection during the test, the approach obtains the average failure frequency of components. After that, failure correlation degree between components is analyzed and the decision algorithm of microreboot group is specified. Research results show that the approach can reboot the system against different failure in componentized distributed applications. As a result, mean time to recovery decreases by about thirty percent.