ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2019, Vol. 56 ›› Issue (4): 779-789.doi: 10.7544/issn1000-1239.2019.20170831

• 系统结构 • 上一篇    下一篇



  1. (西北工业大学计算机学院 西安 710129) (工信部大数据存储与管理重点实验室(西北工业大学) 西安 710129) (
  • 出版日期: 2019-04-01
  • 基金资助: 

Virtual Machine Resources Allocation Methods Based on History Data

Wang Haitao, Li Zhanhuai, Zhang Xiao, Bu Hailong, Kong Lanxin, Zhao Xiaonan   

  1. (School of Computer Science, Northwestern Polytechnical University, Xi’an 710129) (Key Laboratory of Big Data Storage and Management (Northwestern Polytechnical University), Ministry of Industry and Information Technology, Xi’an 710129)
  • Online: 2019-04-01

摘要: 云数据中心中广泛应用虚拟化技术以实现资源的按需分配,从而减小运营成本,提高数据中心的灵活性和可扩展性.然而,虚拟化技术的这些特性也带来了如何在保证虚拟机按需分配的同时,充分利用物理资源而又减小资源冲突率的问题.针对这个问题提出了2种基于历史负载数据的虚拟机资源分配方法,并结合常用的虚拟机放置策略,与现有的常用虚拟机资源分配方法进行对比分析.同时,针对现有的独立评价指标具有片面性的问题,提出1个综合有效性指标,能够结合虚拟机的分配所消耗的物理机数量、物理机的资源利用率以及资源冲突率3方面的指标来综合评价方案的有效性.最后通过实际的云计算负载测试,证明了提出的基于历史数据的虚拟机资源分配方法整体上优于常用的虚拟机资源分配方法,并且综合有效性指标能够合理地从整体上评估虚拟机分配方案的有效性.

关键词: 云数据中心, 虚拟机, 资源分配, 历史数据, 统计分析, 放置策略, 综合有效性指标

Abstract: Virtualization technology is widely used in cloud datacenters to realize on-demand resources allocation so as to lower operating costs. Moreover, the technology can also improve the flexibility and scalability of datacenters. Despite various merits, these features of virtualization technology also introduce an issue about how to allocate the virtual machines to make the best of physical resources while reducing the resource collision rate in the meantime. To this end, this paper proposes two resource allocation methods for virtual machines based on statistical analysis of history data. Combined with commonly-used placement strategies, these two methods are more effective compared with some state-of-art virtual machine resource allocation methods. In addition, existing independent indicators are incomplete to reflect the overall effectiveness of allocation methods. In order to solve the issue, this paper also proposes an integrated effectiveness indicator which combines different indicators from three separate aspects including the number of consumed physical machines, resource utilization and resource collision of physical machines to evaluate the effectiveness of allocation schemes. In the end, through tests of realistic cloud computing overhead, we prove that the proposed allocation methods of virtual machines are superior to common methods, and the integrated effectiveness indicator can reasonably evaluate the overall effectiveness of virtual machine allocation schemes.

Key words: cloud datacenter, virtual machine, resource allocation, history data, statistical analysis, placement strategy, integrated effectiveness indicator