基于OpenStack的云平台调度算法设计与实现

发布时间：2018-05-25 05:13

本文选题：云平台 + 虚拟机　；参考：《北京邮电大学》2014年硕士论文

【摘要】：随着互联网的迅速发展,云计算技术在包括我国在内的世界各国得到广泛的重视、研究和推进。作为一种新兴的计算模式,云计算具有高可靠性、按需分配及动态分配等优点。为了满足应用集中部署的要求,减少硬件成本和维护开销,.电信运营商也开展了云平台的建设。运营商应用系统的资源需求往往会随着业务的发展和用户群的壮大而逐渐增加。然而根据云平台按需分配的基本特点,在创建虚拟机时云平台为系统提供的资源是有限的,难以满足应用资源需求的增长。为了解决这个问题,常见的IaaS平台如OpenStack、VMware均提供动态迁移的功能,将应用迁移到能提供更多资源的虚拟机上,然而电信运营商的一些应用系统要求实时不间断运行,在迁移过程中不可避免的时间开销将会给业务带来影响,造成不必要的损失。因此在将业务应用作为虚拟机部署到云平台时,需要针对电信运营商业务应用场景的特点设计云平台调度算法,在充分考虑减少未来可能的跨物理机迁移次数的同时,以提升平台整体资源利用率为目标,.将虚拟机部署到合适的物理机上。针对此问题,本论文基于教育部-中国移动科研基金2012年度项目“面向互联网的业务支撑系统关键技术及方案研究”,基于OpenStack开源IaaS平台,设计并实现了两阶段云平台虚拟机调度算法。首先,对电信运营商业务进行抽象,研究基于业务规模和生命周期的业务分类,并在此基础上提出两阶段虚拟机资源调度算法；其次,研究OpenStack调度器的实现,并基于OpenStack该实现开发两阶段调度算法模块,用于对算法进行性能评估。最后,对两阶段调度算法及相应模块进行测试,通过对测试结果的分析,验证了调度模块的可用性和两阶段调度算法的有效性。
[Abstract]:With the rapid development of the Internet, cloud computing technology has received extensive attention, research and promotion in the world, including China. As a new computing model, cloud computing has the advantages of high reliability, on-demand allocation and dynamic allocation. In order to meet the requirements of centralized application deployment, hardware costs and maintenance costs are reduced. Telecom operators also carried out the construction of cloud platform. The resource demand of operator application system will increase with the development of service and the expansion of user group. However, according to the basic characteristics of cloud platform allocation according to demand, the resources provided by cloud platform to the system are limited when the virtual machine is created, and it is difficult to meet the increasing demand of application resources. In order to solve this problem, common IaaS platforms such as OpenStackor VMware provide the function of dynamic migration and migrate applications to virtual machines that can provide more resources. However, some application systems of telecom operators require real-time and uninterrupted operation. The inevitable time cost in the migration process will have an impact on the business, resulting in unnecessary losses. Therefore, when deploying the service application as a virtual machine to the cloud platform, we need to design a cloud platform scheduling algorithm according to the characteristics of the service application scenario of telecom operators, which can reduce the number of possible migration times across physical machines in the future. In order to improve the overall resource utilization of the platform as a goal. Deploy the virtual machine to the appropriate physical machine. In order to solve this problem, this paper is based on the project "key Technologies and Solutions of Internet-Oriented Business support system" of the Ministry of Education and China Mobile Research Foundation in 2012, and based on OpenStack open source IaaS platform. A two-stage cloud platform virtual machine scheduling algorithm is designed and implemented. First of all, the service of telecom operators is abstracted, the service classification based on service scale and life cycle is studied, and then a two-stage virtual machine resource scheduling algorithm is proposed. Secondly, the implementation of OpenStack scheduler is studied. A two-stage scheduling algorithm module based on OpenStack is developed to evaluate the performance of the algorithm. Finally, the two-stage scheduling algorithm and its corresponding modules are tested. The availability of the scheduling module and the effectiveness of the two-stage scheduling algorithm are verified by the analysis of the test results.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01

【参考文献】