Research And Implementation Of Job Management System

The wide use of Cluster computer system brings imperative demand for Job Management System (or Cluster Management System). JMS is adopted by many large research organizations and corporations for an important warranty of improving production efficiency and promoting ratio of resource usage. And lots of research and development has been carried out according to the actual need. However, the domestic research in this field has not achieved success. Because of this reason, the JMS is studied in detail. Based on the NQS (Network Queue System), the frame of system and the method of implementation are analyzed and expounded detailedly. In JMS, load balancing is one of the important factors that influence the parallel performance. How to arrange and attempter job rationally, make good use of the disposal ability of nodes and shorten the response time of programs are problems that must be dealing with. Based on analyzing and expounding a good many factors that influence and restrict the efficiency of JMS, a new arithmetic for the JMS is brought forward found on Cluster: Rendezvous Dynamic Load Balancing Arithmetic. The basic thought is that, after evaluating the load of nodes in the Cluster, it flows between the matching nodes according to rule of extreme matching, which makes the load of nodes smooth re-distribution in the Cluster. With simulant experiments, the arithmetic is proved to improve and promote the performance of the system effectively. Additionally, Cluster technique is also used in JMS as an effective exploration. For JMS, a corresponding policy is brought forward. And the system frame of ARM (Active Recovery Manager) is constructed, which promotes the reliability, expansibility and anti- calamity.