通信学报 ›› 2019, Vol. 40 ›› Issue (5): 180-191.doi: 10.11959/j.issn.1000-436x.2019105

• 学术通信 • 上一篇    下一篇

大规模云计算服务器优化调度问题的最优二元交换算法研究

王万良1,臧泽林1(),陈国棋1,屠杭垚1,王宇乐1,陆琳彦2   

  1. 1 浙江工业大学计算机科学与技术学院,浙江 杭州 310027
    2 伦敦大学国王学院工程科学学院,英国 伦敦 WC2R 2LS
  • 修回日期:2019-04-19 出版日期:2019-05-25 发布日期:2019-05-30
  • 作者简介:王万良(1957-),男,江苏高邮人,博士,浙江工业大学教授,主要研究方向为人工智能、运筹优化、大数据。|臧泽林(1994-),男,河北秦皇岛人,浙江工业大学硕士生,主要研究方向为人工智能、运筹优化、深度学习。|陈国棋(1996-),男,浙江温州人,浙江工业大学硕士生,主要研究方向为深度学习、增强学习。|屠杭垚(1996- ),男,浙江杭州人,浙江工业大学硕士生,主要研究方向为智能算法、人工智能、多目标优化。|王宇乐(1990- ),男,江苏常州人,浙江工业大学博士生,主要研究方向为演化计算、多目标优化。|陆琳彦(1996- ),女,浙江杭州人,伦敦大学国王学院硕士生,主要研究方向为通信网络技术。
  • 基金资助:
    国家自然科学基金资助项目(61873240);国家自然科学基金资助项目(51875524)

Research on optimal two element exchange algorithm for large scale cloud computing server scheduling problem

Wanliang WANG1,Zelin ZANG1(),Guoqi CHEN1,Hangyao TU1,Yule WANG1,Linyan LU2   

  1. 1 School of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310027,China
    2 Faculty of Natural and Mathematical Sciences,King’s College London,London WC2R 2LS,England
  • Revised:2019-04-19 Online:2019-05-25 Published:2019-05-30
  • Supported by:
    The National Natural Science Foundation of China(61873240);The National Natural Science Foundation of China(51875524)

摘要:

随着云计算产业的不断兴盛,云计算服务器的合理管理与科学调度成为了一个重要的课题。在模型方面,提出了一个新的携带亲和约束与反亲和约束的混合整数规划(MIP)模型,并将其用于描述大规模云计算服务器调度问题。考虑到求解大规模MIP问题的时间成本,在分枝定界法与局部搜索算法的基础上提出了最优二元交换算法。该算法通过不断地从完整的调度问题中提取MIP子问题,并使用分支定界法解决该子问题的思想,不断地对服务器调度方案进行优化,从而使调度方案接近最优解。实验结果表明,所提算法在测试数据集ALISS上与其他方法相比有较大优势,在完成相同任务的情况下,可以使云计算中心的资源消耗减少4%以上。

关键词: 服务器调度, 混合整数规划模型, 最优二元交换算法, 云计算

Abstract:

With the flourishing of cloud computing industry,the rational management and scientific scheduling of cloud computing servers has become an important issue.In terms of model,a new mixed integer programming (MIP) model with affinity constraints and anti-affinity constraints was proposed to describe the scheduling problem of large scale cloud computing server.Considering the time cost of solving large-scale MIP problems,an optimal two element exchange algorithm was designed with the basics of branch and bound method and local search algorithm.By constantly extracting MIP sub-problems from completing scheduling problems and using branch and bound method to solve the sub-problems,the algorithm continuously optimized the server scheduling schemes,so that the scheduling schemes approached the optimal solution.The experimental results show that the algorithm has great advantages over the other methods in testing data set ALISS,and can reduce the resource consumption of cloud computing center by more than 4% when the same task is completed.

Key words: server scheduling, mixed integer programming model, optimal two element exchange algorithm, cloud computing

中图分类号: 

No Suggested Reading articles found!