| Computer systems have been broadly used in all walks of life since the third wave of industrial revolution brought forth by the computer technologies. These systems have greatly promoted the development of economic society. As a result, more demanding requirements of the reliability and stability of computer systems have been put forward. For example, the server systems in financial, communicational and governmental fields are heavily charged with information processing tasks such as handling, storing and delivering. Once the server systems are out of order, huge losses may be caused. Therefore, it's a burning issue to guarantee the continuous and stable running of key server systems in computer applications.High-availability cluster system aims at providing the end users with continuous and stable services they want and is now a very hot topic in current studies. But as far as the domestic market is concerned, the cluster systems are mostly of the dual-machine backup type, namely, two server systems with one or two RAIDs added. For the time being, when many fields need some key services with high availability, such systems are very costly and very hard to manage. Fortunately the so-called MultiNode Cluster technology has now been developed to meet these requirements and therefore has a very promising market prospect. But due to such factors as technical barriers and investment in development, the high-end market of multi-node cluster systems is essentially owned by foreign manufacturers.In the paper a solution of multi-node system architecture based on storage network is given after considerable analysis by the author. Several core issues and technical difficulties in the realization of the model have been analyzed in a detailed manner. A multi-node system with high availability and basic functions needed, is finally realized.The paper analyzes the market requirements of the high-availability system and the current situation of the uses of cluster technologies. Multi-node cluster systems and their market prospect are given. Design solutions of such system are also studied in the paper. The core of cluster software, namely, cooperation techniques among computers, is paid great attention. By analyzing the existing problems in the heartbeat model of voting, a solution is given to solve the redundancy in linking choice. Heartbeat model is thus established based on multiple data links. The model is capable of choosing proper communication channel, acquiring state information of each node promptly, monitoring the failures of the system accurately and determining the nodes out of use in time. The realization of the model effectively cut down the complexity of coordination among nodes. So the timely and reliable performance is information exchange among nodes is made possible. Also deeply studied in the paper is the failure transferring techniques. Several possible modes are designed. Based on the Markov model, the practicability of the mode"more for 1"in cluster system is proved.The cluster system realized in the paper supersedes the popular dual machine system in many ways. It is technically superior and adapted to the market requirements. The design solution given here is mainly used among users of all walks of life with key services. Continuous operation is guaranteed. The integration of several service environments can thus be obtained. The solution has been accepted by a lot of users from different fields and successfully implemented in the user environment with satisfying results. |
PDF Full Text Request