The Key Technology And Implementation Of Monitoring System For Large Scale Cluster

Currently the trend of high-performance super computer is growing formhigh-end large computer to commercial computer cluster. The reason is the greatdevelopment of PC, workstation and high-speed network system. This make theperformance of cluster system is increasing speedily. So it is possible that the clustertakes place the large computer system. The maintenance and management of cluster ishard because of incompact structure, strongly independent nodes and complex network.In order to solve this problem, the popular method is to build a cluster monitoringsystem above the node operating system. Based on the research of the dawning cluster,analyzes the key technology of large scale cluster monitoring, and design a system oflarge scale cluster monitoring named DCMM2.The structure of DCMM2 is combining hardware and software. DCMM2 is clientand server mode, can monitor the larger scale cluster real time. DCMM2 can gather thestatus of the switch and raid in the cluster system. DCMM2 provides the function ofquerying and analyzing of history data. We can get the alarm of cluster throughmulti-method. We also can control the nodes to power off, power on and reset throughInternetrt. Because DCMM2 use three layers communication model, the delay ofDCMM2 is very short. That make the performance of DCMM2 is excellence. Theabundant function and stability performance of DCMM2, grant the large cluster systemoperating orderly under the efficiency monitoring and controlling. First this paper talk about the state of cluster monitoring, analyzes the methods ofcluster monitoring nowadays, summarizes the merits and shortcoming of these clustermonitoring system. Based on this, we got the goal of DCMM2. DCMM2 is a solvingplan of larger cluster monitoring. After analyzing the key technology of DCMM2,software and hardware combining structure, monitoring cluster resource based onSNMP model, three layers communication model, and communication through smartsocket. Then talking about the whole design plan of DCMM2. Following, validatingthe function and performance of DCMM2. At the end, talking about the problem andnext work of DCMM2.