Research On Key Techniques Of Single System Image Cluster Monitoring System

As the trend of automating system management, cluster system management is on the road to be autonomic. A new type of cluster monitoring system chiefly characterized by single system image (SSI) cluster monitoring, acts as the feedback mechanism in autonomic cluster management system and thus plays an important role in the autonomic management cycle. This dissertation discusses the characters and problems of SSI cluster monitoring system from a technical view, analyzes their corresponding technical objects and challenges, constructs evaluation forces for pondering the efficacy of the related techniques,and goes into detail about the implementation techniques for the kernel of SSI cluster monitoring system. Finally, a SSI cluster monitoring system developed by us, SunEye, is introduced.The characters and problems of SSI cluster monitoring system are closely linked to the specialities of autonomic cluster management system. On the ground of state of the art and the mechanism of autonomic cluster management system, this dissertation analyzes the technical issues of SSI cluster monitoring system from several aspects, including technical objects, technical challenges and evaluation forces. Especially, the analysis outcome of the former two aspects are used to construct the evaluation forces, which are helpful to the analysis and evaluation of the SSI cluster monitoring system related techniques.The kernel of SSI cluster monitoring system consists of information probing module (IPM) and information collection module (ICM). Implementation techniques of IPM and ICM are divided into serverl catalogues, including IPM information probing, IPM syncronization, IPM preprocess, ICM topology, ICM communication model and ICM storage model. This dissertation analyzes the mechanism and application characters of existing techniques,proposes some new techniques and pondering their efficacy and performance. These works give the guidelines for the implementation of SSI cluster monitoring system.At the end of this dissertation, it introduces a SSI cluster monitoring system developed by us, SunEye, which successfully applied to iCluster 1800, a cluster with 256 nodes,and presented high performance.