Adaptive Failure Detector Based on Failure Distribution in P2P Networks

How to quickly find node failures with lower loads on the assumption of guaranteeing the accuracy is the design goal of the failure detector.In this paper,an improved algorithm of the GA-FD detector named F-AFD is proposed.Firstly,the nodes' failure distribution in the P2P system is analyzed.Based on the prediction of the nodes' online probability,the dynamic adjustment to the detection period is realized,which has reduced the detection time's restriction to the detection period.Secondly,an F-AFD detector is proposed,which can provide QoS of failure detection for multi-applications according to quantitative QoS metrics with lower detection loads.Finally,the experimental results validate the correctness and performance of F-AFD.In addition,it proves F-AFD implements a failure detector that belongs to ◇P in the partially synchronous model,which meets the ability to solve fundamental issues such as consensus,atomic broadcast,etc.