面向网格中间件GOS3的OAR集群管理系统

摘要

中国国家网格(China National Grid,CNGrid)是聚合了高性能计算和事务处理能力的新一代信息基础设施的试验床.中国国家网格武汉结点高性能集群系统采用支持资源预留的开源集群管理器OAR,为用户提供具有服务质量保障的高性能计算服务.GOS(Grid Operating System)是CNCrid网格中间件,CNGrideye是CNGrid监控系统.设计了CNGrideye/OAR监控桥接和GOS-HPCG/OAR作业管理桥接机制,使得CNGrideye支持对OAR的监控,GOS HPCG能通过OAR提交作业,充分利用OAR的资源预留等特性.GOS-HPCG/OAR进行基于QoS的作业调度,提高了作业的运行效率和资源的服务性能,达到整合资源的目的.%The CNGrid is a testbed for the new generation of information infrastructure by integrating high performance computing and process transaction capacity.Wuhan site of CNGrid adopts OAR, which is an open-source cluster manager supporting resource reservation,to provide QoS-guaranteed high performance computing services.GOS is the CNGrid support middleware and CNGrideye is the CNGrid monitoring system.This paper designs a CNGrideye/OAR monitor bridging and GOS-HPCG/OAR job management bridging mechanism.Through the above bridging mechanisms, CNGgrideye can monitor the grid system with OAR and GOS HPCG can submit jobs via OAR and exploit the resource reservation capacity.GOS-HPCG/ OAR exploits the QoS-based job scheduling policy to improve the efficiency of job execution and resource utilization and achieve the purpose of integration of resources.