基于Xen的虚拟机资源动态优化算法研究及应用

发布时间：2018-06-28 00:39

本文选题：Xen + 虚拟化　；参考：《东北大学》2013年硕士论文

【摘要】：近年来,虚拟化技术以其能够很好的屏蔽底层硬件资源的异构性,具有提高资源使用效率,增强系统弹性和可扩展性,减少硬件投资及管理维护成本等诸多优势而被广泛的应用于资源服务整合、系统安全、分布式计算以及新兴的云计算等领域。然而,各个虚拟机在运行过程中总会出现资源利用率过高或过低的情况,虚拟化平台中各个物理节点也会出现负载不均衡的现象。在深入分析Xen体系结构的基础上,对虚拟机的CPU和内存资源重新分配以及基于虚拟机动态迁移技术的物理节点负载均衡进行了深入研究,具体做了如下工作。首先,对Xen虚拟化技术进行了深入地研究,针对Xen中资源静态分配而造成的资源利用效率低、分配不合理等问题,提出了一种细粒度的资源调整算法。该算法通过对单个物理节点上各个虚拟机资源使用情况的实时统计分析,从资源利用率低的虚拟机中回收资源,并为利用率较高的虚拟机动态增加资源分配量,该算法能够有效解决资源利用效率低、资源分配不合理的问题,同时也避免了不必要的虚拟机迁移。其次,针对Xen虚拟化平台中各个物理节点上存在负载不均衡的问题,提出了以虚拟机为粒度的粗粒度迁移调度算法,该算法主要包括待迁移虚拟机选择、迁移目标映射和迁移执行三个部分。待迁移虚拟机选择主要从迁移代价和迁移后对物理机负载的影响两方面来选择最适合的迁移对象；在标准粒子群算法的基础上,重新设计了能够满足多种资源均衡的多目标适应度函数,并改进了粒子之间的距离计算公式,改进后的粒子群算法能够有效的寻找待迁移虚拟机与迁移目标宿主机之间的映射关系：迁移执行则是利用迁移冲突检测来寻找最优的、可并行的迁移执行顺序,以减少总迁移时间。第三,设计实现了一个云环境下的虚拟机资源动态优化框架,提出了热点检测策略,并结合了上述两种资源动态优化算法进行热点消除,提高了虚拟化平台的资源利用效率和用户服务质量,降低了管理和维护成本。以及解决了粗细两种粒度优化算法可能存在的冲突问题。最后,通过大量实验对所提出的资源动态优化算法进行了验证。实验结果表明资源动态优化算法能够有效解决资源分配不合理和负载均衡等问题。
[Abstract]:In recent years, because of its ability to shield the heterogeneity of underlying hardware resources, virtualization technology can improve the efficiency of resource use, enhance the flexibility and expansibility of the system. It is widely used in the fields of resource service integration, system security, distributed computing and new cloud computing due to the advantages of reducing the cost of hardware investment and management and maintenance. However, every virtual machine always has too high or too low resource utilization in the running process, and each physical node in the virtualization platform will also appear the phenomenon of load imbalance. Based on the in-depth analysis of Xen architecture, the CPU and memory resource redistribution of virtual machines and the load balancing of physical nodes based on virtual machine dynamic migration technology are studied. Firstly, the Xen virtualization technology is deeply studied, and a fine-grained resource adjustment algorithm is proposed to solve the problems of low resource utilization efficiency and unreasonable allocation caused by static resource allocation in Xen. Through the real-time statistical analysis of each virtual machine resource on a single physical node, the algorithm can recover the resource from the virtual machine with low resource utilization, and dynamically increase the resource allocation for the high utilization virtual machine. The algorithm can effectively solve the problems of low efficiency and unreasonable allocation of resources and avoid unnecessary migration of virtual machine. Secondly, aiming at the problem of load imbalance on each physical node in Xen virtualization platform, a coarse-grained migration scheduling algorithm based on virtual machine granularity is proposed, which mainly includes the selection of virtual machine to be migrated. Migration target mapping and migration execution are three parts. The selection of virtual machine to be migrated mainly from the migration cost and the impact on the physical machine load to select the most suitable migration object; on the basis of standard particle swarm optimization algorithm, The multi-objective fitness function, which can satisfy the multi-resource equilibrium, is redesigned, and the formula for calculating the distance between particles is improved. The improved particle swarm optimization algorithm can effectively find the mapping relationship between the virtual machine to be migrated and the host: migration execution uses migration conflict detection to find the optimal parallel migration execution order. To reduce the total migration time. Thirdly, we design and implement a virtual machine resource dynamic optimization framework under cloud environment, propose a hot spot detection strategy, and combine the above two resource dynamic optimization algorithms to eliminate the hot spot. It improves the resource utilization efficiency and user service quality of virtualization platform, and reduces the cost of management and maintenance. And the conflict between the two granularity optimization algorithms is solved. Finally, the proposed resource dynamic optimization algorithm is verified by a large number of experiments. The experimental results show that the dynamic resource optimization algorithm can effectively solve the problems of unreasonable resource allocation and load balancing.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP302

【相似文献】