基于OPENSTACK的天线仿真云平台构建及其资源调度算法优化

发布时间：2018-03-15 01:18

本文选题：天线仿真云　切入点：OpenStack　出处：《天津工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着天线仿真技术的发展和智能天线的普及,高校、中小企业等研究机构对计算机软硬件等基础设施的需求随之快速增长,频繁的设备更新和相对独立的软、硬件运行模式,导致了天线仿真的成本逐年攀升。随着云计算和虚拟化技术的兴起,可以将天线仿真与云计算技术相结合,打破制约天线仿真技术发展的瓶颈。在高校或中小企业建设实验室环境下的天线仿真云平台,不仅能够使计算机软硬件资源得到合理利用,而且可以极大地方便科研人员的实践活动,有效提升科研效率。本文首先介绍了几种主流的IaaS云平台,并选择了被业界广泛接受的OpenStack技术构建了天线仿真云平台,创建了天线仿真云平台必备的仿真环境模板文件和云主机类型,为用户提供了按需使用的仿真环境和计算机硬件的服务;然后深入研究了 OpenStack的整体架构和核心组件,结合OpenStack源码,对OpenStack的初始虚拟机分配模块、虚拟机动态迁移模块进行了深入地分析,阐述了其工作原理,并指出其中存在的不足之处;最后,针对OpenStack原生资源调度策略存在的缺陷,分别提出了多目标蚁群优化算法和虚拟机动态迁移多目标优化算法。利用多目标蚁群优化算法对虚拟机初始放置策略进行改进,通过信息素的持续更新快速地获取到最优解,从而为新建的虚拟机找到最佳的放置位置。仿真结果表明,该算法既可以保证良好的服务性能,又能够降低资源负载及电量损耗,确保数据中心达到一个良好的运行状态。利用虚拟机动态迁移多目标优化算法对OpenStack虚拟机的动态迁移策略进行改进。设计了二分上整延时法及时间预测法确定迁移时机,利用多目标优化算法选取恰当的目标物理主机,设计了基于概率的选择算法用以规避虚拟机的群聚效应。CloudSim仿真结果表明,这一整套虚拟机资源动态调度方法能够胜任对资源进行实时调度的工作,同时能够优化数据中心的性能。
[Abstract]:With the development of antenna simulation technology and the popularization of smart antenna, the demand for computer hardware and software infrastructure in universities, small and medium-sized enterprises and other research institutions has increased rapidly, frequent equipment updates and relatively independent software and hardware operation mode, With the rise of cloud computing and virtualization technology, antenna simulation can be combined with cloud computing technology. The antenna simulation cloud platform can not only make rational use of computer software and hardware resources, but also break the bottleneck restricting the development of antenna simulation technology. Moreover, it can greatly facilitate the practical activities of researchers and effectively improve the efficiency of scientific research. Firstly, this paper introduces several mainstream IaaS cloud platforms, and selects the widely accepted OpenStack technology to construct antenna simulation cloud platform. The necessary simulation environment template file and cloud host type of antenna simulation cloud platform are created, which provide users with the simulation environment and computer hardware service on demand, and then deeply study the overall architecture and core components of OpenStack. Combined with OpenStack source code, the initial virtual machine allocation module and virtual machine dynamic migration module of OpenStack are deeply analyzed, its working principle is expounded, and the shortcomings are pointed out. Aiming at the defects of the native resource scheduling strategy of OpenStack, the multi-objective ant colony optimization algorithm and the virtual machine dynamic migration multi-objective optimization algorithm are proposed, and the multi-objective ant colony optimization algorithm is used to improve the initial placement strategy of virtual machine. Through the continuous updating of pheromone, the optimal solution can be obtained quickly, so as to find the best placement position for the new virtual machine. The simulation results show that the algorithm can not only guarantee good service performance, but also reduce the resource load and power consumption. The dynamic migration strategy of OpenStack virtual machine is improved by using the multi-objective optimization algorithm of virtual machine dynamic migration. The binary integral delay method and time prediction method are designed to determine the migration opportunity. Using the multi-objective optimization algorithm to select the appropriate target physical host, the probability-based selection algorithm is designed to avoid the clustering effect of virtual machine. The simulation results show that, This set of virtual machine resource dynamic scheduling methods can perform the task of real-time resource scheduling and optimize the performance of the data center at the same time.
【学位授予单位】：天津工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN820;TP391.9

【参考文献】