关键词: 自律计算，性能故障，阂值检测，自我监视，策略

Abstract: Autonomic computing system often uses threshold to mark performance faults. Threshold is the boundary value of performance counter,which will be detected automatically by the system. When the performance counter of one of the devices reaches the threshold, it is considered that performance fault occurs on one of the devices. Then the autonomic manager will actively select policies to recover these faults,which therefore makes the whole system to maintain normal state. Nowadays,it lacks of research on system performance faults in autonomic computing field. On the basis of studying the self-monitoring of autonomic computing system, this paper proposed a multiple data threshold detecting method,by detecting the exceeding and recovering of the threshold in multiple times, system can effectively judge whether the performance fault occurs. This method assures the detecting accuracy,which provides the effective basis to recover fault for autonomic computing system.

Key words: Autonomic computing, Performance faulting, Threshold detecting, Self-monitoring, Policies