异构云环境下能耗感知的虚拟机网络优化调度研究

发布时间：2018-04-23 08:39

本文选题：异构云环境 + 能耗感知　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：云计算技术环境下的数据中心大多逐步采用虚拟化技术向云服务模式迁移,从而使得运行在异构云环境上的虚拟机网络日益复杂。为了满足用户不断增长的云计算需求,作为重要载体的云数据中心数量和规模也日益的庞大,云数据中心运营商面临着高能耗、高成本等严峻问题;而在异构云环境下,虚拟机到物理服务器映射关系的差异会导致数据中心能源消耗的不同,使得能耗感知的虚拟机网络优化调度问题成为云数据中心资源管理的重要问题。云数据中心管理者面对着如何度量好整个数据中心的能源消耗,感知云数据中心的能耗情况,从而对异构云数据中心的能耗进行高效的管理;如何实现用最少的能源消耗来满足数据中心虚拟机网络的放置与运行;如何根据虚拟机在运行时的动态变化及时调整整个数据中心的运行情况,以降低数据中心的能耗,从而减少服务成本等相关问题。为此,本文对异构云环境下能耗感知的虚拟机网络优化调度进行了研究。首先,针对异构云数据中心的能耗感知问题,本文将采用适用于异构云环境下能源消耗的度量方法;建立异构云数据中心能源消耗模型,其中包括虚拟机能耗模型,物理机能耗模型,网络设备能耗模型等;测量与预测异构云环境下的基础设施能耗,为研究能耗感知的虚拟机网络优化放置和调度提供支持。其次,针对能耗感知的虚拟机网络优化放置问题,本文将对虚拟机网络优化放置问题建立问题模型;并分别采用免疫遗传算法和基于最小割与最佳适应的改进算法对虚拟机网络放置问题进行求解;通过与现有数据中心中常见的放置算法以及一些启发式算法进行对比,分析免疫遗传算法和基于最小割与最佳适应的改进算法对虚拟机网络放置问题的求解质量。再次,针对能耗感知的虚拟机网络动态迁移问题,本文将对通过统计用户的历史访问习惯,分析用户行为特征;进一步给出用户行为特征模型并通过虚拟机负载预测建立特征模型;给出基于用户行为分析的虚拟机网络动态迁移算法的问题描述、问题模型和算法步骤,并通过对比实验进行分析。最后,在虚拟机网络优化放置和动态迁移的理论研究基础上,对能耗感知的云数据中心资源管理平台进行设计和开发。设计能耗感知的云数据中心资源管理平台的架构;对各功能模块、数据库、关键算法部分进行设计;完成平台两大中心模块的实现。
[Abstract]:Most data centers in cloud computing environment gradually adopt virtualization technology to migrate to cloud service mode, which makes virtual machine network running in heterogeneous cloud environment more and more complex. In order to meet the increasing demand of cloud computing, the number and scale of cloud data centers, which are important carriers, are increasingly huge. Cloud data center operators are faced with severe problems such as high energy consumption and high cost. The difference of mapping relationship between virtual machine and physical server will lead to different energy consumption of data center, which makes the optimal scheduling of virtual machine network aware of energy consumption become an important problem in resource management of cloud data center. Cloud data center managers are faced with how to measure the energy consumption of the whole data center and perceive the energy consumption of the cloud data center so as to efficiently manage the energy consumption of heterogeneous cloud data center. How to achieve the minimum energy consumption to meet the data center virtual machine network placement and operation, according to the dynamic changes of the virtual machine in the run time adjust the entire data center operation, in order to reduce the energy consumption of the data center, To reduce service costs and other related issues. Therefore, this paper studies the optimal scheduling of energy-aware virtual machine networks in heterogeneous cloud environments. First of all, aiming at the problem of energy consumption perception of heterogeneous cloud data center, this paper will adopt the energy consumption measurement method suitable for heterogeneous cloud environment, and establish the energy consumption model of heterogeneous cloud data center, including virtual machine energy consumption model. Physical computer energy consumption model, network equipment energy consumption model and so on; measure and predict the energy consumption of infrastructure in heterogeneous cloud environment, which provides support for the research of virtual machine network optimization placement and scheduling. Secondly, aiming at the problem of optimal placement of virtual machine network based on energy consumption awareness, this paper establishes a problem model for optimal placement of virtual machine network. Immune genetic algorithm (IGA) and improved algorithm based on minimum cut and optimal adaptation are used to solve the problem of virtual machine network placement, which is compared with common placement algorithms and some heuristic algorithms in existing data centers. The quality of solving virtual machine network placement problem based on immune genetic algorithm and improved algorithm based on minimum cut and optimal adaptation is analyzed. Thirdly, aiming at the problem of dynamic migration of virtual machine network with energy consumption awareness, this paper will analyze the characteristics of user behavior through statistics of users' historical access habits. Furthermore, the user behavior feature model is given and the feature model is established through the virtual machine load prediction, and the problem description, problem model and algorithm steps of the virtual machine network dynamic migration algorithm based on user behavior analysis are given. And through the contrast experiment carries on the analysis. Finally, based on the theoretical research of virtual machine network optimization and dynamic migration, the energy-aware cloud data center resource management platform is designed and developed. Design the structure of energy consumption aware cloud data center resource management platform; design each function module, database, key algorithm part; complete the implementation of two central modules of the platform.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP302;TP393.09

【参考文献】