多目标优化的虚拟机调度模型与关键算法研究

发布时间：2019-02-12 11:28

【摘要】：虚拟化技术是构建云计算环境的技术基石之一,特别是服务器虚拟化技术,在基础设施层为云计算技术提供了强有力的实现手段。目前,数据中心作为构建云计算资源池的基本载体,通过虚拟化的方式以物理机服务单元承载虚拟机,同时以虚拟机服务单元来承载应用服务。云环境下,面对数据中心中物理资源和虚拟资源的大量聚集,如何通过虚拟化资源的分派和调度改变传统IT的资源提供与管理模式,实现IT资源的集约共享,提高资源整体利用率,降低服务成本,并保证服务质量是当前迫切需要解决及改善的重要问题。本文首先分析了目前关于虚拟机调度相关技术的前沿研究状况,并总结了当前研究所存在的问题；通过对虚拟机调度目标与相关要素之间关系的分析,提出了多目标优化的虚拟机调度模型,系统总结了与虚拟机调度目标相关的决定因素、次要因素、影响范围以及所涉及的基本要素,定性和定量探究了基本要素间的内在联系和相互作用的边界条件；在此基础上进而建立了衡量虚拟机资源分派和调度有效性的指标体系；进而提出基于傅立叶变换的资源分派算法和基于资源需求特征匹配的虚拟机调度算法,最后通过实验验证算法的有效性。本文的主要创新点如下： 1)厘清了虚拟机资源调度目标与决定调度的基本要素之间的关系,提出了多目标优化的虚拟机调度模型； 2)构建了衡量虚拟机资源调度有效性的指标体系； 3)基于数据中心整体的资消耗需求规律以及虚拟机资源需求的数据特点,提出基于傅立叶变换的资源分派算法,确定了数据中心为满足应用负载所需要的最佳开机数量序列以及各虚拟机资源配置序列； 4)建立了虚拟机性能敏感度表达以区分虚拟机类型,并在此基础上提出了基于资源需求特征匹配的虚拟机放置算法。
[Abstract]:Virtualization technology is one of the technological cornerstones of building cloud computing environment, especially server virtualization technology, which provides a powerful means for cloud computing technology in the infrastructure layer. At present, as the basic carrier of constructing cloud computing resource pool, data center hosts virtual machine as physical machine service unit and application service unit as virtual machine service unit through virtualization. In the cloud environment, in the face of a large number of physical and virtual resources gathering in the data center, how to change the traditional IT resource supply and management mode through the allocation and scheduling of virtualized resources to realize the intensive sharing of IT resources. Improving the overall utilization of resources, reducing the cost of service, and ensuring the quality of service is an urgent need to solve and improve the important issues. Firstly, this paper analyzes the current research status of virtual machine scheduling related technologies, and summarizes the problems existing in the current research. Based on the analysis of the relationship between virtual machine scheduling objectives and related elements, a multi-objective optimization model of virtual machine scheduling is proposed, and the determinants and secondary factors related to virtual machine scheduling goals are summarized systematically. The scope of influence and the basic elements involved, qualitative and quantitative exploration of the inherent relationship between the basic elements and the boundary conditions of interaction; On this basis, an index system is established to measure the efficiency of virtual machine resource allocation and scheduling. Then a resource allocation algorithm based on Fourier transform and a virtual machine scheduling algorithm based on resource requirement feature matching are proposed. Finally, the effectiveness of the algorithm is verified by experiments. The main innovations of this paper are as follows: 1) the relationship between virtual machine resource scheduling objectives and the basic elements determining scheduling is clarified, and a multi-objective optimized virtual machine scheduling model is proposed. 2) constructing the index system to measure the efficiency of virtual machine resource scheduling; 3) based on the rule of capital consumption and the data characteristics of virtual machine resource demand, a resource allocation algorithm based on Fourier transform is proposed. The optimal number of boot sequences and each virtual machine resource configuration sequence for the data center to meet the application load are determined. 4) the performance sensitivity expression of virtual machine is established to distinguish the type of virtual machine, and a virtual machine placement algorithm based on resource requirement feature matching is proposed.
【学位授予单位】：复旦大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP302

【相似文献】