虚拟化环境下的内存节能机制研究

发布时间：2018-01-09 22:03

本文关键词：虚拟化环境下的内存节能机制研究　出处：《华中科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：在当前的大型数据中心中，虚拟化技术以其良好的资源整合和安全隔离等特性被广泛使用。随着大容量物理内存的使用，虚拟化环境下的内存能耗已经逐渐成为数据中心中最主要的能耗产出者。当前虚拟化环境下内存节能技术并未考虑虚拟机物理内存的分布对其能耗的影响，从而导致节能效果不佳。当前物理内存包含很多可控单元（Rank），每个Rank包含不同的能耗状态。虚拟机在动态创建和使用的过程中，其物理内存分布到很多不同的Rank中，因此当该虚拟机运行时，，多个Rank处于高能耗状态，从而大大增加了其物理内存能耗。针对上述问题，分别在静态环境下和动态环境下提出并实现了基于内存节能的虚拟机内存分配系统，从虚拟机内存分配的角度减少虚拟机运行时占用的物理块数目，从而减少其内存能耗。该系统包含两种主要的机制，一是在虚拟机管理器（VMM）中重新设计了内存页管理的方法，以更加简单和高效的方式组织内存页，从而方便上层的内存操作；二是在该内存页管理方法的基础上针对虚拟机内存和VMM内核堆内存从内存节能的角度设计了不同的启发式内存分配算法。该算法尽量高效的将请求的物理内存页分配到尽量少的Rank中，使得系统运行时更多的Rank处于低能耗状态，从而减少总的物理内存能耗。实验表明，改进后的内存分配系统跟原有的内存分配系统相比，静态环境下内存能耗节省了近50%，动态环境下内存能耗节省超过60%。该数据表明改进后的内存分配系统极大的降低了虚拟机化环境下的物理内存能耗。
[Abstract]:In current large data centers, virtualization technology is widely used for its good resource integration and security isolation, with the use of large physical memory. Memory energy consumption in virtualized environment has gradually become the most important energy output in the data center. In the current virtualization environment, memory energy saving technology does not consider the impact of virtual machine physical memory distribution on its energy consumption. The current physical memory contains many controllable units, each Rank contains a different state of energy consumption. Virtual machines are dynamically created and used. Its physical memory is distributed to many different Rank, so when the virtual machine is running, many Rank are in high energy consumption state, which greatly increases its physical memory energy consumption. Aiming at the above problems, a memory allocation system based on memory saving is proposed and implemented in static environment and dynamic environment respectively. From the point of view of virtual machine memory allocation, the memory consumption of virtual machine is reduced by reducing the number of physical blocks. The system contains two main mechanisms. First, the method of memory page management is redesigned in virtual machine manager (VMMM) to organize the memory page in a simpler and more efficient way, so as to facilitate the memory operation of the upper layer. The second is to design different heuristic memory allocation algorithms from memory saving point of view for virtual machine memory and VMM kernel heap memory based on this memory page management method. Save pages are allocated to as few Rank as possible. It makes more Rank in low energy state when the system is running, thus reducing the total energy consumption of physical memory. The experimental results show that compared with the original memory allocation system, the improved memory allocation system saves nearly 50% memory energy in static environment. The data show that the improved memory allocation system greatly reduces the physical memory energy consumption in the virtual machine environment.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333.1

【共引文献】