The high performance computing cluster in the Northwest Regional Meteorological Center uses the job scheduling software LoadLeveler to schedule and manage jobs, in order to realize the maximization of resource utilization and throughput. This paper studies and discusses the role and method of scheduling management job. Defining the queue through the configuration file keywords, setting the scheduling algorithm and scheduler adjustment parameters, matching the processing requirements of the job with the available resources, 〖JP2〗the availability rules of clusters and various resources are defined according to actual requirements, so as to prevent the mutual interference between jobs. Combined with the LoadLeveler job preemption function, the regional numerical weather prediction business model is guaranteed to have the priority and stable operation. Perl language is used to write a script to account the resources of the high performance computer cluster, so that the administrator can fully grasp the use of resources of various application modes, reasonably adjust the system resource allocation and scheduling strategy, and give full play to the benefits of the high performance computer cluster in the Northwest Regional Meteorological Center.〖JP〗