虚拟化系统中处理器资源管理和控制技术研究

发布时间：2018-10-21 19:48

【摘要】：资源的有效利用一直是虚拟化技术的研究重点，而处理器资源作为虚拟化系统中的核心资源，对它的有效管理和使用将显著提升虚拟化系统的性能随着芯片集成度的逐步提高，多核处理器的核心数目不断增多，，提供了丰富的计算资源硬件辅助虚拟化技术的引入，也使处理器资源具有了诸多新的特性在当前硬件基础条件下，虚拟化系统中处理器资源的有效管理和使用面临新的挑战各种动态管理和控制策略的引入，完善了虚拟机处理器资源管理的手段和方法，并取得了较好的效果但目前的研究仍存在以下问题：1）多核处理器资源仍未得到充分利用2）虚拟机调度算法不能有效适应负载特性3）用户实时订制所需的虚拟机执行环境比较困难针对上述问题，提出了一种多层次的处理器资源动态配置模型，将资源粒度划分为物理处理器VCPU和VCPU运行时间三类，引入了多层次多粒度的资源动态调整策略，满足了虚拟机处理器资源的自主动态按需调整在分析虚拟机负载特性的基础上，设计了一种基于负载特性的虚拟机调度算法LCVS，该算法可以保证在不破坏公平调度的原则下，提高I/O受限型虚拟机的响应速度基于Intel VT技术设计了一种动态配置虚拟机执行环境的机制，能够使用户按需定制虚拟机执行环境，改变虚拟机的运行时特性，满足了特定背景下的用户需求通过对测试结果进行分析，得出结论：动态配置模型相比静态配置模型可以有效提升虚拟机性能；设置调整观察区间可以使调整更加及时准确；LCVS算法可以缩短I/O受限型虚拟机50%~60%的响应时间，同时较好的平衡了CPU受限型虚拟机和I/O受限型虚拟机对处理器的使用；相比于传统方式，虚拟机执行环境动态配置机制增强了VMM的可用性
[Abstract]:The effective utilization of resources is always the focus of the research of virtualization technology. As the core resources of virtualization system, the effective management and use of processor resources will significantly improve the performance of virtualization system with the gradual improvement of chip integration. The increasing number of cores of multi-core processors provides rich computing resources for the introduction of hardware-assisted virtualization technology, and also makes processor resources have many new features under the current hardware infrastructure. The effective management and use of processor resources in virtualization systems face new challenges. The introduction of various dynamic management and control strategies has improved the means and methods of virtual machine processor resource management. Good results have been achieved but the following problems still exist in the current research: 1) the multi-core processor resources have not been fully utilized 2) the virtual machine scheduling algorithm can not effectively adapt to the load characteristic 3) the virtual machine needed by the user to customize in real time. The implementation environment is more difficult to address the above problems, In this paper, a multi-level processor resource dynamic configuration model is proposed, which divides the resource granularity into physical processor VCPU and VCPU runtime, and introduces a multi-level multi-granularity resource dynamic adjustment strategy. On the basis of analyzing the load characteristics of virtual machine, a virtual machine scheduling algorithm (LCVS,) based on load characteristics is designed, which can ensure that the virtual machine scheduling algorithm can not destroy the fair scheduling principle. To improve the response speed of the I / O restricted virtual machine, a mechanism for dynamically configuring the execution environment of the virtual machine is designed based on Intel VT technology, which enables the user to customize the execution environment of the virtual machine on demand and change the runtime characteristics of the virtual machine. Through the analysis of the test results, it is concluded that the dynamic configuration model can effectively improve the performance of virtual machine compared with static configuration model. The LCVS algorithm can shorten the response time of 50% of the I / O limited virtual machine and balance the use of the CPU limited virtual machine and the I / O limited virtual machine to the processor. Compared with traditional methods, dynamic configuration mechanism of virtual machine execution environment enhances the usability of VMM
【学位授予单位】：解放军信息工程大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP332

【相似文献】