期刊文献+

存储系统中多访问流的特性分析

Analyzing Integrated Workload Characteristics of Multi-streams on Storage Systems
下载PDF
导出
摘要 研究多用户读写流的特性是存储性能研究的重要课题,尤其是IP-SAN使得共享SAN的用户数大大增加。首次理论分析了多IO流的综合特性。考察了多IO流在访问特性和到达特性上的相互影响。重点分析了由于IO流访问的局部性与多IO流的交叉到达而大大增加了的系统寻址动作。这些寻址动作影响着多IO流请求的服务时间,以及设备的整体吞吐量。同时,还开发了一个配置灵活的负载模拟器IOsim,验证了它的有效性和精确性。利用IOsim验证了多IO流特性的理论分析。模拟试验的结果证明了理论分析的正确性。使用提出的理论分析方法能够有效预测并判断多IO流并发访问的各种综合特性以及性能变化趋势。最后描述了未来的工作。 Analyzing integrated workload characteristics of multi-streams on storage systems is important for evaluating storage performance. For example, the number of users sharing SAN storage becomes greater in IP-SAN than before. The integrated workload characteristics of multi-streams in theory for the first time is analyzed , inclu- ding mutual influence between these various streams in access pattern and arrival pattern. Especially seeking be al- so have investigated address that effects service time of I0 requests and throughput of storage system because of I0 stream accessing locality and multi-streams' staggered accessing. A workload simulator named IOsim been also have implemented that is configurable flexibly, and validating its accuracy and validity. Then using IOsim, the multi-streams' characteristics obtained from theory analyzing and simulating to prove that the theory analyzing ac- cords with simulation very well is compared. So the set of formulabe can used to predict and estimate multi- streams' various characteristics and the change
出处 《科学技术与工程》 北大核心 2012年第31期8287-8292,8312,共7页 Science Technology and Engineering
基金 国家"八六三"高技术研究发展计划基金项目(2002AA112010 2002AA1Z2102B)资助
关键词 存储系统 IO流特性 局部性 负载模拟器 storage systems IO stream trend of storage system performance as I0 streams change. characteristics locality workload simulator IOsim
  • 相关文献

参考文献14

  • 1Ganger G R. Generating representative wynthetic workloads. In : Pro-ceedings of 21st International Computer Measurement Group Confer-ence, Nashville, TN, 1995, 1263-1269.
  • 2Zhu Y L, Zhu S Y, Xiong H. Performance analysis and testing of thestorage area network. In : Proceedings of 10th NASA Goddard Confer-ence on Mass Storage Systems and Technologies,Maryland,2002 :108-120.
  • 3Shriver E, Merchant A, Wilkes J. An analytic behavior model fordisk drives with readahead caches and request reordering. In: Pro-ceedings of ACM SIGMETRICS Conference on Measurement andModeling of Computer Systems, Ontario, Canada, 1998:182-191.
  • 4Kuratti A, Sanders W H. Performance analysis of the RAID 5 diskarrays. In: Proceedings of international computer performance anddependability symposium, Erlangen, Germany, 1995:236-245.
  • 5Mishra S K, Mohapatra P. Performance study of RAID - 5 disk arrayswith data and parity cache. In : Proceedings of the International Con-ference on parallel processing, Bloomingdale, 1996:222~229.
  • 6Thomasian A,Menon J. Approximate analysis for fork-join synchroni-zation in RAID5. Computer systems : Science and Engineering,1997;12(5) : 329-338.
  • 7Varki E, Merchant A, Xu J, et al. An integrated performance modelof disk arrays. In: the 11th International Workshop on Modeling, A-nalysis and Simulation of Computer Telecommunications Systems,Or-lando, FL, 2003 :296-305.
  • 8Varki E,Merchant A,Xu J,et al. Issues and challenges in the per-formance analysis of real disk arrays. IEEE Transactions on Paralleland Distributed Systems, 2004; 15(6) ; 559-574.
  • 9Gomez M E, Santonja V. Analysis of self-similarity in I/O workloadusing structural modeling. In: the 7th International Workshop onModeling, Analysis and Simulation of Computer and Telecommuni-cation Systems, Maryland, 1999:97-104.
  • 10Wang Mengzhi, Tara M,Chan N H,ef al. Data mining meets per-formance evaluation: fast algorityms for modeling bursty traffic. In:Proceedings of 18 th international conference on data engineering,California, USA, 2002:507-516.

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部