Design And Implementation Of Server Cluster Monitoring System

Current as the network bandwidth constantly widen and the price of computer continues to fall, pc cluster, composed by general pc as basic computing node, with the advantage of high performance, cost-effective, good extensibility and other advantages, has taken placed the traditional supercomputer and widely used in many industrial fields. Each node in the cluster is connected by high-speed network and the cluster relies on the collaboration of software and hardware to gain its computability. As the number of node is growing and single node failure is easily, system monitor plays an important role in using the cluster effectively. By monitoring, we can see the resource usage of each node easily; locate the failure node; analysis operational tread, performance limits and bottlenecks of whole cluster, providing evidence for task scheduling, and administrator’s management.This project comes from the author’s graduate internship, which aims to monitor the health of server cluster in data center of company. The system’s feature includes: collect data about the state of each node’s various components, such as node’s load, CPU usage, memory usage, network traffic; processing the monitor data and push it into database and notify the user about abnormal data according personal pre-defined rules made by user; visual display the data to the authorized terminal users according to access rule in the form of web page for query and future processing. In this paper,/proc virtual file system was used to collect monitor data, XML was used for data transmission, DB Proxy was used to build high-performance, highly available MySQL database cluster as a storage component. Both C/S architecture and B/S architecture are used in system, composed by four components:the agent distributed on each node, the proxy on several node, monitor server, and data storage. In the process of design and implementation of the system, the author mainly completed data collecting of monitoring agent, monitoring configuration management, monitoring data visualization, user and privilege management of monitoring server, in the early and late stage of the project, participated in the requirement research and testing of system.The monitoring system can monitor the data center effectively, has taken little impact on the performance of cluster. The agent distributed on each node take little resource.