Research Of Failure Detection Technology Based On Self-Organized Neighborhood

Accompanying with the development of computer network technology, some newdistributed computing systems have been gradually widely applied. The requirements oflarge-scale distributed system have become even higher in many critical domains. Inorder to improve the quality of distributed system services, it not only requires thatsystem quickly accomplish the relevant applications, but also requires the system havethe characteristics of high availability, high scalability, low cost. Especially for thesenodes which provide network services in the complex network environment, how toensure the reliability and availability of them has become very important, therefore, itrequires the introduction of node failure detection technology. This technology lays thefoundation for the further works, which could provide the service of fault tolerance andfailure recovery.However, traditional failure detection methods hardly meet the requirements ofhigh scalability and self adaptability. When these methods confront the network trendsof large-scale, dynamic, complexity, they will make the frequent occurrence of thephenomenon, such as local network congestion, increase of transmission delay andmisjudgment rate. These factors directly make the fault detection efficiency decline.Based on the existing research works, we proposed the dynamic failure detectionmethod based on self-organized neighborhood, the details of research work are asfollows:①The obstacles of P2P networking technology were surveyed and analyzed,according to the real network environment, in this paper, the NAT traversal schemebased on UDP was designed and implemented, it could make the nodes located indifferent LAN to successfully use P2P communication, and according to theself-organized neighborhood technology, the failure detection model was designed.Then through the hierarchical self-organizing classification method, the large scaletarget network was divided into many smaller autonomous domains, eventually thesystem’s scalability and adaptability were improved.②Aim at the problems of high communication cost and long transmission time,in this paper, the P2P logic overlay network based on physical topology was researched,then the pretreatment method based on clustering IP address and the neighborhoodpartitioning method based on network distance were proposed, finally the P2P logical topology of overlay network were highly corresponded with real physical topology, theefficiency of communication was improved.③The communication requirements of different nodes, the efficient messagetransmission method was designed and implemented, and the self-adaptive Gossiprevised mechanism based on path records was proposed, the dissemination efficiencyof message intercommunication in group was improved.④Aim at the drawbacks of traditional failure detection methods, the predictionmethod based on exponential smoothing was analyzed and researched, and the staticand dynamic detection mode were combined, then the new failure detection methodwith multimode was proposed, finally it could fit dynamic nature of networkenvironment, and the phenomenon of misjudgment rate rising which caused by networkcongestion and transient node invalid were reduced.To sum up, in this paper, the key problems which caused by node failure detectionmethod applied on the new distributed network environment were analyzed, then therelated nodes management based on self-organized neighborhood and messagesdissemination methods were analyzed and designed. Finally the whole dynamic failuredetection model was implemented, and its high efficiency was proved.