Cluster Monitoring System Research And Design

Nowaday cluster technology is becoming an active research area in the field of high-performance parallel computing. Applying the cluster technology to system not only can shorten the response time of system, but can improve usability, reliability and scalability of system. At present the relevant research of cluster is still scanty, but it already has aroused the attention of people, so it is very essential to study the relevant technology of cluster.Monitor system is very important to management to cluster as core,the data which Monitor system get can be used to balance the load,assigned task to every node and adjust relatively service to system.beside those,it can report to administrator about the problems such as toubles or disables of hardware or software.which can help the improve the reliabily of cluster.but to cluster,beside get more infomations about system,what more,it need to do with this problem properly.fix it,by tracing timely task.First, considering the information processing of command and control system as background, this thesis analyzes the information processing, system architecture and characteristic of cluster. Based on these, a general architecture of cluster is presented, and several problems such as operation, clock synchronization, fault-tolerant are analyzed too.Next, according to the architecture of cluster, the architecture of Monitor system, the evaluating method of system's performance, the mechanism of decision and response are discussed separately. At last the information collecting, perturbations elimination, event association and failure removal are discussed deeply.In Cluster environment,the number of nodes is not content number,it frequcently add or deduce it nodes according what it needs,so that problem which the cluster will meet are different from each other, therefore timely management is adopted to cluster to add or delete node and change their propertys. As to the troubles and defaults,throught abandent design plan to recovery to reach a high availiary relialiy.based this presume conditions, try to design a minitor system which distribution the system ,with high efficiency,easy to extentsion feature.At last, to whole Monitor and control system, architecture is integrated plan to us,by this ,we can deploy every function modules which do some work for the system,some function modules can be put together,or some can fulfill its job indepence,the most important point is how them work tightly,especially to our system,monitoring system,task balancing,and some other data relatively task,how they can do their job by one or more plan or by some arichtecture,here we will discussion this problem .Especilly to LVS,nowday it is a pop load balance structure,it is a good example for this...