R&D Of Network Job Management System

With high-powered microprocessor and high-speed computer network coming, it is rather simple than possible to connect computer system consisting of a large number of CPU by high-speed network. Network job management system is such kind of middle software that founds upon the operating system, and manages the workstations or PCs of the cluster. The author of this dissertation has researched this system for a long time, after that, considering the high-speed development of the Internet application and the requirement of the user, she developed the WEB client of this system. During the course of developing, considering the heavy load of the www server itself~ the author brought forward to divide the whole system to two parts, the communicating module named orbclient is the design stress, which communicates with job management system server and all the CGI programs. To implement the communication between this module and the CGI programs, a set of detailed communicating protocol has been designed. At the same time, during the course of developing the project, the author analyzed and compared all the design pattern of the service programs, and selected the optimized pattern to implement the system.Cluster is founded upon the distributed system, and it is the direction to resolve the high availability, high reliability and scalability problems. For example, like the WEB client project the author developed, the users?request come from anywhere of the WAN environment at anytime, in order to ensure the project抯 high efficiency and robust, the cluster technology can be used. Cluster technology resolved the loadbalancing problem by distributing the request dynamically, which means, redirecting the request to the node with the light load, and giving the response to the user with the less time. The failover refers to when a node happened to a failure, another node would take over this fail node, and when the broken-down recovers, still the original node will provide the service. This procedure is transparently to the user. The author understands this potential direction and brings a load-balancing algorithm aim at resolving the network load balance-Q&R algorithm, through the test, gets the result that using this algorithm resolve the network load balance is more effective. At the mean while, the author researches the schemes about the checkpoint set to resolve the failover problem.