Research And Implementation Of The Infiniband Cluster Visualization Monitoring System

Cluster is a set of nodes—computers joined physically by high-performance network or LAN(local area network) and is a concrete means of the stable parallel computing architecture. As a lower-cost, high-performance parallel computer system, it is widely used in parallel processing. The monitoring and control of cluster is an important instrument of mastering how well every node operates and performs. The present development of the system is varied according to various architecture and purposes.Aimed at the current system architecture used frequently when to build a cluster which uses the InfiniBand high-performance network with a bandwidth of 20Gb/s. The article focuses on the following areas:The reasons that bring the interference and the feasible ways to minimize the interference. The reasons for the interference are mainly frequent data gathering and the bandwidth occupied by internet transmission, as well as the intrusion of the illegal uses and the sleeping system service's occupation of the system resources. And the article will put forward feasible ways to improve the situation.It proposes using binary tree architecture to gather monitoring result with speed in a diffuse way, effectively enhancing the efficiency of the gathering. The binary tree's layer architecture helps spread message quickly, and because every node runs parallel, the simultaneous efficiency of information gathering is enhanced. The growing speed of nodes in the binary tree is relatively fast, as a result, with the expansion of the cluster scale, it can display its advantages more effectively.The safety test of the cluster is an important means to ensure the stable and safe running of the cluster against interference. In the text, according to the monitoring system architecture, a method of testing the network safety and the system safety is proposed.At the end, combined with the design and development of the cluster monitoring software CLS, the text researches on the feasibility of the method to decrease the monitoring system's interference in the cluster...