Research On Enhancement Technology Of Proxy Cluster Dependability Based On Autonomic Computing

With the developing of IT technology and the networking of applicationsystems, in order to improve the processing capability of the system, distributedcluster systems are widely applied to various fields. While, Internet securityincidents are more and more frequent, the cluster systems are facing a variety ofsecurity threats at all times. Due to the constraints of distributed cluster systemcomplexity, keeping cluster systems dependable becomes a thorny problem,designing more complex software is a classics traditional approach to achieve thegoal of enhancing the dependability of the cluster system, but it is unable to meetcurrent need. Autonomic computing technology provides new avenues forenhancing cluster system dependability. By using the characteristic of autonomiccomputing self-managing, the system dependability can be enhanced automatically,and autonomic computing avoids the complexity of the system itself. Proxy is abuzzword on the Internet today, as a public-oriented distributed system, proxycluster is faced with massive access, and the dependability problem is also severe.Based property of autonomic computing "self-managing", this thesis proposes amechanism for enhancing the dependability of the proxy cluster.This thesis introduces the autonomic computing technology, proposes proxyclusters dependability self-enhancing framework, then the relationship between themonitor measures are analyzed, ARX model is used to model the linear relationshipbetween the monitoring measures; after that, according the mapping relationshipbetween monitor measures and modules, the method to discover abnormal modulebased on Bayesian classification techniques is proposed. The target of abnormalstate self-awareness is achieved. As to the problem of the optimal action sequencegenerated in the research of dependability self-decision, an algorithm for generatingstrategy automatically is designed; while this thesis researches on the way ofcharacterizing abnormal state by using the fluctuations of monitor measures as aneigenvalue, a one-dimensional approach of the abnormal state characteristic is proposed, and constructs the B+tree Knowledge of dependability enhancementstrategy with the one-dimensional eigenvalue. At last, the Experiment which mainlyfocuses on two dependability threats malicious sessions and cache hot, is carried out.Experiment results show that the mechanism can enhance autonomic dependabilityof proxy cluster effectively.In summary, according the research on enhancement technology of proxycluster dependability based on autonomic computing, this thesis achieves the goal ofproxy cluster abnormal state self-aware and self-decision, provides effective securityprotection for proxy cluster system.