Cluster System Job Scheduling Optimization Technology Research And Implementation

Cluster system has the low cost, high performance characteristic which provides the formidable batch and parallel computation ability and represents the main direction of high performance computer. However, simply improving the performance of hardware can't meet the user's various needs, especially the needs of large-scale science computing and business dealing, highly efficient management of cluster resource is needed. On this demand the cluster job management system appeared and fast to develop. It manages and schedules software and hardware resources according to user demand to guarantee the average share of resources and improve the system utility and output.In spite of competitive management ability, the present popular job management systems are unable to manage and schedule jobs according to various software and hardware resource requirements in different application domains, thus it is needed to be improved. Based on the job management system LSF, this paper has focused research on resource management optimization of cluster system, and realized corresponding scheduling policy.1. Resource management multidimensional QoS model of cluster computing system is proposed based on the abstraction of system resources and user resource requirements.2. Job scheduling algorithm based on Multidimensional QoS(JSBMQ) and resource matching algorithm based on Multidimensional QoS(RMBMQ) applying to different application area are proposed.3. Scheduler mainly using JSBMQ is designed and realized based on job management system LSF. Interrelated functions supporting scheduler are optimized.The research and optimized work in this paper has effectively enhanced scheduling efficiency and system utilization of job management system, at the same time reduced user's job response time and met the QoS requirement in different application domain. So it has the certain actual value.