期刊文献+

基于Backfilling调度算法的“扩履适足”改进算法 被引量:2

An Algorithm for Backfilling-Based "Enlarge Five to Ten"Parallel Job Scheduling
在线阅读 下载PDF
导出
摘要 在众多的并行作业调度算法中,Backfilling通常被广泛认为是有效提高CPU利用率的一种算法。该算法是在FCFS算法的基础上,将队列中较小的作业回填(Backfill)到空闲CPU,以提高CPU利用率。但是,当空闲CPU数量仍然无法满足Backfilling算法中小作业的回填要求时,系统仍有部分CPU闲置,因而也难以达到更好地提高CPU利用率的目的。对于共享内存体系结构的并行计算机系统,本文提出了基于Backfilling算法的“扩履适足”的改进算法。该算法以正在运行的作业的CPU利用率为依据,通过动态调整正在运行作业的CPU数,扩大可供回填(backfill)的CPU空间,使得Backfilling算法无法回填的作业得到运行,弥补了Backfilling算法的不足,大大提高了共享内存体系结构并行计算机系统的CPU利用率。 There is a wide agreement that backfilling produces significant benefits in scheduling parallel jobs among many algorithms. Backfilling is based on the FCFS algorithm, which backfills smaller queued jobs into idle CPU's to improve CPU utilization. But when idle CPUs are still not able to meet the needs for the backfilling algorithm to backfill smaller queued jobs, there are still some CPU's which are idle. So it is difficult to get the goal of improving CPU utilization. As for memoryshared architecture parallel computer systems, this paper presents an "Enlarge Five to Ten" algorithm based on the backfilling algorithm. The algorithm is based on the CPU utilization of running jobs to enlarge the CPU space for backfilling by adjusting dynamically the number of CPU's which are running jobs, so it can backfill the jobs which can't be backfilled by the backfilling algorithm. The presented algorithm remedies the drawbacks of the backfilling algorithm and improves the CPU utilization of the memory-shared architecture parallel computer systems.
出处 《计算机工程与科学》 CSCD 2006年第9期94-96,122,共4页 Computer Engineering & Science
关键词 并行计算 作业调度 CPU利用率 Backfilling算法 扩履适足 parallel computing jobs scheduling CPU utilization backfilling algorithm enlarge five to ten
  • 相关文献

参考文献9

  • 1A W Mu'alem,D G Feitelson.Utilization,Predictability,Workloads,and User Runtime Estimates in Scheduling the IBM SP2 with Backfilling[J].IEEE Trans on Parallel and Distributed System,2001,12(6):529-543.
  • 2D G Feitelson,A Mu'alem Weil.Utilization and Predictability in Scheduling the IBM SP2 with Backfilling[A].12th Int'l Parallel Processing Symp[C].1998.542-546.
  • 3U Schwiegelshohn,R Yahyapour.Improving First-Come-Fir-st-Serve Job Scheduling by Gang Scheduling[A].IPPS'98 Workshop:Job Scheduling Strategies for Parallel Processing(LNCS1459)[C].1998.180-198.
  • 4Saeed Iqbal,Rinku Gupta,Yung-Chin Fang.Planning Considerations for Job Scheduling in HPC Clusters[EB/OL].http://www.dell.com.downloads/global/power/ps1qo5-20040135-Fang.PDF,2005-11.
  • 5Jochen Krallmann,Uwe Schwiegelshohn,Ramin Yahyapour.On the Design and Evaluation of Job Scheduling Algorithms[A].Proc of the Job Scheduling Strategies for Parallel Processing(LNCS1059)[C].1999.17-42.
  • 6D A Lifka.The ANL/IBM SP Scheduling System[A].IPPS'95 Workshop:Job Scheduling Strategies for Parallel Processing(LNCS 949)[C].1995.395-303.
  • 7S Srinivasan,R Kettimuthu,V Subramani,et al.Characterization of Backfilling Strategies for Parallel Job Scheduling[A].Proc of the ICPP Workshop on Scheduling and Resource Management for Cluster Computing[C].2002.514-519.
  • 8Edi Shmueli,Dror G Feitelson.Backfilling with Lookahead to Optimize the Packing of Parallel Jobs[J].Journal of Parallel and Distributed Computing,2005,65(9):(1090-1107).
  • 9D G Feitelson,M A Jette.Improved Utilization and Responsiveness with Gang Scheduling[A].Job Scheduling Strategies for Parallel Processing(LNCS1291)[C].1997.236-261.

同被引文献15

  • 1Foster I. Intemet computing and the emerging grid. Nat- ureWeb Matters[ EB\OL]. ( 2000-124)7 ). http ://www. nature, com/nature/webmatters/grid/grid, html.
  • 2Foster I, Kesselman C. The Grid: BluePrint for a new computing inrfastruetuer [ M ]. San Fransisco: Morgan Kaufmann Publishers Inc, 1999.
  • 3Foster I. What is the Grid? A three point cheeklist[ J]. GRID Today,2002,1 (6) :22-25.
  • 4Wolski R, Spring N, Hayes J. The network weather serv- ice :A distributed resource performance forecasting serv- ice for metacomputing [ J ]. Future Generation Computer Systems, 1999,15 (5/6) :757-768.
  • 5Lifka D. The ANL/IBM SP scheduling system[ C] ffPro- ceedings of JSSPP, 1995 : 295-303.
  • 6Talby D, Feitelson D. Supporting priorities and improving utilization of the ibm sp scheduler using slack based backfilling[ C ] //Proceedings of the 13th International and 10th Symposium on Parallel and Distributed Process- ing, 1999:513-517.
  • 7Wong A K L, Goscinski A M. Evaluating the EASY-back- fill job scheduling of static workloads on clusters [ C ] // IEEE International Conference on Cluster Computing, 2007:64-73.
  • 8Wong A K L, Goscinski A M. The impact of under-esti- mated length of jobs on EASY-Backfill scheduling[ C]// The 16th Euromicro Conference on Parallel, Distributed and Network-Based Processing,2008:343-350.
  • 9Buyya R, Murshed M. GridSim:a toolkit for the modeling and simulation of distributed resource management and scheduling for Grid computing [ J ]. Concurrency and Computation : Practice and Experience, 2002,14 ( 13/14/ 15) :1175-1120.
  • 10Parallel Workload Archive [ EB \ OL ] [ 2010-08-25 ]. www. cs. huji. ac. il/labs/parallel/workload.

引证文献2

二级引证文献2

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