计算机科学 ›› 2015, Vol. 42 ›› Issue (2): 191-197.doi: 10.11896/j.issn.1002-137X.2015.02.041

• 软件与数据库技术 • 上一篇    下一篇

高性能计算中一种改进的数据访问节能技术研究

邓定胜   

  1. 四川民族学院计算机科学系 康定626001
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受四川省教育厅自然科学一般项目:量子程序中一种改进的延时估计算法研究(15ZB0332),四川省教育厅自然科学重点项目:基于粗糙集理论的数据挖掘算法研究(13ZA0136)资助

Research on Improved Energy-saving Technology for Data Access in High Performance Computing

DENG Ding-sheng   

  • Online:2018-11-14 Published:2018-11-14

摘要: 针对现有磁盘功率管理方法无法处理高能耗并行应用的短空闲周期、大规模的代码更改和能耗开销较大等问题,提出了一种面向编译器的数据访问调度技术。该技术包括两个阶段:在第一阶段,编译器分析并行应用程序,提取磁盘访问模式,然后生成调度表;在第二阶段,“数据访问调度器”根据调度表进行数据访问。与先前基于软件的策略相比,所提方法不需要改变代码或数据结构。实验评估结果表明,对于数据密集型工作负载,所提方法可以有效提升节能效果,节能率从5.5%上升到11.8%,从而增加了磁盘降速策略对数据密集型高性能计算的可行性。此外,它也将多速磁盘的节能效果从12.7%增加到了27.6%。

关键词: 磁盘功率管理,空闲周期,代码,数据访问,编译器,能量

Abstract: Aiming at the problems that the existing disk power management methods cannot handle short idle periods of high-performance parallel applications,require extensive code modifications and cost much energy,this paper proposed a compiler-directed data access scheduling technology for saving disk energy.The proposed technology contains two phases.In the first phase,the compiler analyzes parallel application programs,extracts disk access patterns and generates scheduling tables,while in the second phase,a “data access scheduler” performs the actual data accesses according to scheduling tables.As compared to prior software based efforts,our framework requires no code or data restructuring.Our experimental evaluation reveals that the proposed framework effectively increases the energy savings brought by the disk power down mechanism for data-intensive workloads,and the rate of energy savings improves from 5.5% to 11.8%,making disk spin-down a viable strategy in data-intensive high-performance computing.In addition,it increases the energy benefits of multi-speed disks from 12.7% to 27.6%.

Key words: Disk power management,Idle periods,Code,Data access,Compiler,Energy

[1] 易会战,罗兆成.基于动态电压调节的高性能业务系统能耗优化[J].华中科技大学学报:自然科学版,2013(1):25-29
[2] 张凯,陈书明,王耀华,等.面向通用 HPC 的高性能 DSP 设计权衡[J].计算机学报,2013,36(4):790-798
[3] 张帅,宋风龙,王栋,等.多核结构片上网络性能-能耗分析及优化方法[J].计算机学报,2013,36(5):988-1003
[4] Hadjipaschalis I,Poullikkas A,Efthimiou V.Overview of current and future energy storage technologies for electric power applications[J].Renewable and Sustainable Energy Reviews,2009,13(6):1513-1522
[5] Li K,Kumpf R,Horton P,et al.A quantitative analysis of disk drive power management in portable computers[C]∥USENIX winter.2002:279-291
[6] Gurumurthi S,Sivasubramaniam A,Kandemir M,et al.DRPM:dynamic speed control for power management in server class disks[C]∥30th Annual International Symposium on Computer Architecture,2003.IEEE,2003:169-179
[7] Son S W,Chen G,Kandemir M,et al.Exposing disk layout to compiler for reducing energy consumption of parallel disk based systems[C]∥Proceedings of the tenth ACM SIGPLAN symposium on Principles and practice of parallel programming.ACM,2005:174-185
[8] Son S W,Kandemir M,Choudhary A.Software-directed diskpower management for scientific applications[C]∥19th IEEE International Parallel and Distributed Processing Symposium,2005.IEEE,2005:4-13
[9] 李元章,孙志卓,马忠梅,等.S-RAID5:一种适用于顺序数据访问的节能磁盘阵列[J].计算机学报,2013,36(6):1290-1302
[10] 刘靖宇,郑军,李元章,等.混合 S-RAID:一种适于连续数据存储的节能数据布局[J].计算机研究与发展,2013,50(1):37-48
[11] 孟涛,刘浩,胡宏扬.基于预读策略的节能数据访问技术[J].计算机工程,2012,38(8):44-46
[12] 李建敦,彭俊杰,张武.云存储中一种基于布局的虚拟磁盘节能调度方法[J].电子学报,2012,40(11):2247-2254
[13] Weissel A,Beutel B,Bellosa F.Cooperative I/O:A novel I/O semantics for energy-aware applications [J].ACM SIGOPS Opera-ting Systems Review,2002,36(SI):117-129
[14] Papathanasiou A E,Scott M L.Energy efficient prefetching and caching[C]∥Proceedings of the 2004 USENIX Annual Technical Conference.Berkeley,CA,USA,2004:255-268
[15] Pinheiro E,Bianchini R.Energy conservation techniques for disk array-based servers[C]∥Proceedings of the 18th annual international conference on Supercomputing.ACM,2004:68-78
[16] Son S W,Kandemir M,Choudhary A.Software-directed diskpower management for scientific applications[C]∥19th IEEE International Parallel and Distributed Processing Symposium,2005.IEEE,2005:4b-4b
[17] Son S W,Kandemir M.Energy-aware data prefetching for multi-speed disks[C]∥Proceedings of the 3rd Conference on Computing Frontiers.ACM,2006:105-114
[18] Thakur R,Gropp W,Lusk E.Data sieving and collective I/O inROMIO[C]∥The Seventh Symposium on the Frontiers of Massively Parallel Computation,1999(Frontiers’99).IEEE,1999:182-189
[19] Liao W,Coloma K,Choudhary A,et al.Collective caching:application-aware client-side file caching[C]∥14th IEEE International Symposium on High Performance Distributed Computing,2005(HPDC-14).IEEE,2005:81-90
[20] Calder P C,Jacobsen C,Skall Nielsen N,et al.Nutritional benefits of omega-3 fatty acids[M]∥Food enrichment with omega-3 fatty acids.2013:3-26
[21] Whaley J.Joeq:A virtual machine and compiler infrastructure[C]∥Proceedings of the 2003 Workshop on Interpreters,VirtualMachines and Emulators.ACM,2003:58-66
[22] Palmer A J,Brandt A,Gozzoli V,et al.Outline of a diabetes disease management model:principles and applications [J].Diabetes research and clinical practice,2000,50:S47-S56
[23] Ligon W,Ross R.PVFS:Parallel virtual file system[M]∥Beowulf cluster computing with Linux.MIT Press,2001:391-429

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!