An Optimization Approach to Restart Tree Based on Mean Failure Frequencies

Restart tree is the groundwork for microreboot and recursive recovery, while optimization of restart tree is a key problem in increasing efficiency of recursive recovery. Based on the research of related works, the optimization principle of restart tree is analyzed and a novel approach is proposed for optimizing restart tree. In this approach, mean failure frequencies of components are obtained by injecting exceptions during load testing, which could be used to compute and analyze the failure correlation degrees between components. The components with high failure correlation degrees are constructed as a restart group to realize the optimization of restart tree. The results of the case study show that this approach reduces the mean time to recovery of componentized distributed application and implements the optimization of restart tree.