基于数据中心的两阶段虚拟机能效优化部署算法

发布时间：2018-06-01 22:14

本文选题：云计算 + 数据中心　；参考：《计算机应用》2014年11期

【摘要】：针对数据中心在虚拟机动态部署过程中的高能耗问题,提出了面向数据中心的两阶段虚拟机能效优化部署算法——DVMP_VMMA。第一阶段为初始部署,提出了动态虚拟机部署(DVMP)算法限定主机最优部署数量,降低了闲置能耗;同时,为了应对负载的动态变化,第二阶段提出迁移约束的虚拟机迁移算法(VMMA)对初始部署方案作进一步优化,这样不仅得到的系统能耗更低,而且还能保证应用服务质量。与满载算法(FL)、基于固定门限值的部署算法(FT),绝对中位差部署算法(MAD)、四分位差部署算法(QD)、迁移周期最优算法(MTM)、最小占用率迁移算法(MIU)进行的比较实验结果表明:DVMP_VMMA不仅考虑了系统能耗优化,使运行时资源利用率更高;而且还可以避免VM频繁迁移完成对性能的提升,其在优化数据中心能耗、SLA违例、VM迁移量的控制及性能损失等指标上均有较好效果,其综合性能优于对比算法。
[Abstract]:In order to solve the problem of high energy consumption in the dynamic deployment of virtual machines, a two-stage virtual machine energy efficiency optimization algorithm named DVMPVMMA-oriented data center is proposed. The first stage is the initial deployment. A dynamic virtual machine deployment (DVMPP) algorithm is proposed to limit the optimal number of hosts deployed, which reduces the idle energy consumption, and in order to cope with the dynamic changes of the load, In the second stage, a migration constrained virtual machine migration algorithm (VMMA) is proposed to further optimize the initial deployment scheme, which not only reduces the system energy consumption, but also ensures the application quality of service. Compared with full load algorithm, fixed threshold based deployment algorithm, absolute median difference deployment algorithm, quartile difference deployment algorithm, migration cycle optimal algorithm and minimum occupancy migration algorithm, the experimental results show that DVVMPSP VMMA is not good. Only considering the optimization of system energy consumption, It can also avoid VM frequent migration to complete performance improvement, which has good effect on optimizing data center energy consumption SLA-illegal VM migration control and performance loss, etc. Its comprehensive performance is better than the contrast algorithm.
【作者单位】：武汉轻工大学数学与计算机学院;中国科学院云计算中心;
【基金】：武汉轻工大学引进人才科研启动基金资助项目(2014RZ03)
【分类号】：TP308;TP302

【参考文献】