【作者】 黄秋兰； 朱随江； 程耀东； 陈刚；

【Author】 Qiulan Huang, Suijiang Zhu, Yaodong Chen, Gang Chen (Institute of High Energy Physics,Chinese Academy of sciences,100049)

【机构】 中国科学院高能物理研究所计算中心；

【摘要】 GRASS(Grid-enabled Advanced Storage System)分级存储系统主要负责提供海量数据的分级存储管理和数据迁移功能。目前,GRASS系统的数据迁移仅包含存储池水位、数据迁移时间间隔等系统级别的策略,而缺乏针对文件的精细控制。针对该问题,本文设计了GRASS文件预留系统,实现基于用户级的数据预留,目的是为了高能物理的研究人员可以根据需要将自己指定的文件预留在磁盘上,从而避免自己要访问的文件被迁移到磁带库中去,从而提高用户访问数据的速度。本文首先介绍了GRASS系统,并给出了GRASS文件预留系统的基本架构,接着详细阐述系统的设计和实现,最后给出了系统的运行结果及测试分析。更多还原

【Abstract】 GFRS(GRASS File Reservation System) is designed to improve the file access performance of GRASS(Grid-enabled Advanced Storage System) which is a Hierarchical Storage Management(HSM) system developed at Computing Center, Institute of High Energy Physics. GRASS can provide massive storage management and data migration, but the data migration policy is simply based factors such as pool water level, the intervals for migration and so on, so it is short of precise control over files. As for that, we design GFRS to implement user-based file reservation which is to reserve and keep the required files on disks for High Energy physicists. CFRS can improve file access speed for users by avoiding migrating frequently accessed files to tapes. In this paper we first give a brief introduction of GRASS system and then detailed architecture and implementation of GFRS. Experiments results from GFRS have shown good performance and a simple analysis is made based on it.更多还原