The Study Of High Availability Of Network Servers Cluster System And Management Software Realization

With the explosive growth of commercial service via Internet, High Availability of server system is strongly required for the benefits of servers. In addition, the current server systems have to be scaled to meet the development of service content and the diversity of service. The practical high availability system in use cannot satisfy the scalable server system, so it is important to do some research on scalable, high-available cluster system.The advantage of cluster system is scalability. While the nodes in cluster become larger, the reliability of the whole system will be lower. The clients of different services on server cluster system are so big that high availability means much of server cluster. High-availability software is an essential part of cluster system. Fault detection, fault recovery and management software is especially important for large scale and long running application. The framework of cluster makes high availability possible while server cluster system make high availability essential. Research on combination of high availability and cluster is an important branch of computer field in recent years.In this paper we review the state of the art techniques including server cluster's architecture and required performance of server cluster software. Then based on the fundamentals of high availability and key techniques, we do some research and improvement on double dispatcher high availability system. Then we induce a new technique named software rejuvenation in cluster high availability system, implement a high availability management system, give some models and algorithms of software aging and proposed a new algorithm. At last, we give an evaluation of cluster high availability system.