高能物理研究所校园网格计算环境设计研究

摘要

自1996年VAX小型机退役后,已建立了几套PC-Farm系统,它们分别为不同的物理实验做数据分析所使用,运行不同的作业管理系统如NQS、PBS和Condor等.由于各个实验数据分析对计算系统使用的时间不同,致使整个系统的使用效率不高.为了使这些分立运行的系统发挥更好的效用,使用网格计算技术将这些分立的计算资源整合起来,形成一个计算能力更强大的计算系统,以便这些实验组能够在更大范围内对计算资源共享.给出了校园网格计算环境的硬件配置、体系结构等设计方案,以及各个组件的功能等.%There have been several PC-Farm computing systems built for various experiments at IHEP, since VAX machines were retired in 1996.These PC-Farms are running separately with different job management systems such as NQS or PBS, this leads to making inefficient use of these clusters. In order to share these PC-Farm computing systems across IHEP,the implementation and deployment of Globus and Condor system, and the sharing policy among different groups, the developing Web-based user′s job submission interface are described. For storage, CASTOR is used as HSM for backup/restoring software, therefore, there are CEs and only one SE in the moment at IHEP.