The Research And Implementation Of Highly Scalable Cluster Management System Based On Web Service

With the rapid development of computer technology, the cluster system came into being. With its enormous computing power, security and mass storage, high scalability, dynamic stretching, and low cost, cluster system has been widely used in all walks of life. The powerful functions of cluster need to be based on the good management. The management of cluster will have a direct impact on the quality of services provided by cluster system, so build a safe, easy to use, stable cluster management system is very important.A cluster is a large-scale computing system, and its huge node amount and heterogeneity both bring big challenges to its management and monitoring. The growth in nodes quantity, the ceaseless emerging of new applications and the continuous development of cluster technology proposed new requirements of Cluster management. How to make the system easy to use, easy to maintain and easy to manage, how to ensure that all nodes work efficiently, detect changes in node status in time and quickly respond, to make the cluster system work steadily and continuously, how to improve the system scalability solving the bottleneck caused by the system size increasing, making system functions easy expand, applying new technology to meet the needs of new applications, all of these requirements make the focus and difficulty of cluster system management task.This paper designed and implemented an extensible cluster management system based on WEB service after an in-depth study of cluster management system in domestic and international and a deep thought of characteristics of computer cluster applications in school. Using cloning technology, the system achieved function of remote installation and restoration, resolving the problem that it is difficult to install and deploy a cluster. Using the command channel, the single command interface, the system realized the remote resource manager and distributed command console tool and designed and realized cluster node auto-discovery mechanism, making the cluster easy to be used, the system designed the alarm and action mechanism and realized an remote task manager and an alarm of cluster node using WMI, IPMI technology, solving the problem that the cluster system is difficult to monitor. The system took many measures to solve the problem of cluster scalability:three-tier management structure to solve the problem of the expansion of the system; WEB services of basic cluster management to support new application development; the "modular, scalable request processing framework" and "extensible command frame" to realize the expansion and the optimized hot swappable character of the node functions.