基于能效的数据中心资源整合机制

发布时间：2018-02-20 02:42

本文关键词： 云计算数据中心动态迁移资源整合绿色能效 OpenStack　出处：《电子科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着人类社会的飞速发展，相应的环境问题也愈加突出，因全球变暖而引发的极端天气不断向人们发出警告，绿色能效、降低碳排放已成为研究热点，这一理念也深刻影响着IT领域。数据中心自发展之初就面临巨大的电能需求，而随着虚拟化技术的逐步升级，云计算理念的深入人心，利用“瘦前端”接入“云端”服务器直接使用架设在数据中心上的系统功能摆脱硬件配置束缚已经不再是设想，利用虚拟技术中的动态迁移技术取代传统业务整合为绿色能效资源管理问题开辟了新的技术途径。本文第一章首先简单介绍了云计算以及数据中心的历史背景，介绍了相关技术的发展路线，并提出了基于能效的数据中心资源整合机制问题。其次，整合机制的设计与实现都需要有实时监控数据的支持，，在第二章中本文分析介绍了三种常用的数据中心监控系统，并针对其数据中心电源监控/管理和虚拟机监控扩展做了分析与概括，最后提出两种现阶段性能较优的数据中心监控解决方案。再次，本文第三章详细介绍了现阶段数据中心的能耗问题，介绍了两种基于实时监控数据的服务器能耗模型，简要说明了现有硬件节能技术的功能与局限，并基于该能耗模型提出了物理机能效整合的基本准则。随后提出了基于双阈值的能效整合机制及整合机制的三个核心问题：阈值设置，待迁移虚拟机选择，虚拟机资源整合机制，最后本章提出三种虚拟机选择机制以供选择。第四，为了更详细的分析资源整合机制本文将基于能效的资源管理问题中的虚拟机资源整合问题单独提出来，在第四章针对多虚拟机迁移场景从多种不同角度进行了分析和研究，提出了三种整合机制。RP整合机制基于随机放置，存在不可控性；FFD整合机制将迁移整合转化为装箱问题，但忽略了源主机的区别，导致算法存在不稳定性；GCBFD整合机制将分组思想引入FFD整合机制，提供了更好的算法性能和算法稳定性。仿真结果表明，GCBFD整合机制稳定准确且有效的降低了云数据中心的能耗，为最优选项。最后，在第五章中，本文基于OpenStack开源云平台进行了资源整合机制的开发设计，设计了基于OpenStack的能耗感知型资源整合智慧云平台的模块功能和逻辑接口。
[Abstract]:With the rapid development of human society, the corresponding environmental problems become more and more prominent. The extreme weather caused by global warming constantly warns people. Green energy efficiency and reducing carbon emissions have become the research focus. The data center has been facing huge power demand since the beginning of its development, but with the gradual upgrading of virtualization technology, cloud computing concept has been deeply rooted in people's hearts and minds. It is no longer envisaged to use the "thin front end" to access the "cloud" server directly using the system functions set up in the data center to get rid of the constraints of hardware configuration. The use of dynamic migration technology in virtual technology to replace traditional business integration has opened up a new technical approach to green energy efficiency resource management. In the first chapter of this paper, the historical background of cloud computing and data center is briefly introduced, the development route of related technology is introduced, and the resource integration mechanism of data center based on energy efficiency is put forward. Secondly, the design and implementation of the integration mechanism need the support of real-time monitoring data. In the second chapter, this paper analyzes and introduces three kinds of commonly used data center monitoring system. The data center power monitoring / management and virtual machine monitoring extension are analyzed and summarized. Finally, two kinds of data center monitoring solutions with better performance at present are put forward. Thirdly, the third chapter introduces the energy consumption of the data center in detail, introduces two kinds of server energy consumption models based on real-time monitoring data, briefly explains the function and limitation of the existing hardware energy-saving technology. Based on the energy consumption model, the basic principles of energy efficiency integration of physical machines are proposed. Then, the energy efficiency integration mechanism based on double thresholds and three key issues of integration mechanism are proposed: threshold setting, choice of virtual machine to be migrated, Finally, three mechanisms of virtual machine selection are proposed. 4th, in order to analyze the mechanism of resource integration in more detail, this paper puts forward the problem of virtual machine resource integration in resource management based on energy efficiency. In Chapter 4th, we analyze and study the multi-virtual machine migration scenarios from different angles, and propose three kinds of integration mechanisms .RP integration mechanism based on random placement, there is an uncontrolled FFD integration mechanism to transform migration integration into packing problem. However, the differences between the source host and the source host are ignored, which leads to the instability of the algorithm and the introduction of the grouping idea into the FFD integration mechanism. The simulation results show that the GCBFD integration mechanism is stable, accurate and effective to reduce the energy consumption of the cloud data center and is the best option. Finally, in chapter 5th, we design the resource integration mechanism based on OpenStack open source cloud platform, and design the module function and logic interface of energy consumption aware resource integration intelligent cloud platform based on OpenStack.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP308

【参考文献】