【作者】 颜田； 赵祥虎； 马震太； 张晓梅； 李卫东；

【Author】 YAN Tian;ZHAO Xianghu;MA Zhengtai;ZHANG Xiaomei;LI Weidong;Institute of High Energy Physics,Chinese Academy of Science;

【机构】 中国科学院高能物理研究所；

【摘要】 高能物理实验产生的数据量十分巨大,离线数据处理需要大量的计算资源。同时,高能物理实验通常是由分散在世界各地的大学和研究机构共同参与完成的,合作单位间有分享数据的需求。在此背景下,分布式计算系统得到了广泛的应用。高能所为BESIII实验建立了一套基于DIRAC中间件的分布式计算系统,经过Multi-VO扩展,目前同时支持BESIII,CEPC,JUNO等多个实验。本文研究了一种基于StoRM+Lustre的网格存储模型,描述了它的原理、优点,以及对它的性能测试和分析,最后介绍了它在DIRAC分布式计算系统中的应用。StoRM+Lustre存储将标准的网格存储和大规模并行文件系统Lustre相结合,实现了本地大容量存储的网格化,以及网格存储对分布式计算用户的透明化,方便了站点间数据共享和用户对分布式计算的使用。更多还原

【Abstract】 The data produced by modern high energy physics(HEP experiment is very huge. Massive computing resources are needed by offline data processing. At the same time, HEP experiment usually accomplished by many institute and universities worldwide. They wish to share the experiments data among others. The distributed computing is widely used under this background. In Institute of High Energy Physics(IHEP) at China, we build a distributed computing system for BESIII experiment, this system is based on DIRAC middleware. After multi-VO extension, it supports several experiments, like BESIII, CEPC, JUNO. In this paper, we study a storage model based on StoRM+Lustre. It’s principles, advantages is described, and the performance test and its analysis is presented. The application of StoRM+Lustre in IHEP’s distributed computing environment is demonstrated too. The StoRM+Lustre storage integrates standard grid storage and massive parallel file system Lustre. It makes the local storage grid-aware, and make grid storage transparent to end user. The data sharing between sites are benefit from it. It also makes the distributed computing more convenient to physics users.更多还原