基于OpenLava的虚拟机调度子系统的设计与实现

发布时间：2018-01-23 12:40

本文关键词： 集群 Open Lava KVM 作业调度虚拟化　出处：《西安科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着高性能计算的发展,计算机集群系统以其卓越的性价比、良好的可扩展性和高可用性,日益成为当今高性能计算机体系结构的热点问题。在计算机集群系统中,借助集群作业调度系统将集群中分布的资源统一管理,根据作业的资源需求合理分配资源,充分利用集群计算资源,提高系统的资源利用率。本文基于OpenLava集群作业调度系统,结合用户实际需求,设计并实现了虚拟机调度子系统,解决了Linux高性能集群中由于缺少Windows计算节点,而导致Windows应用作业不能执行的问题,实现了针对不同的系统应用作业提供合适的计算节点资源的目的,提高集群现有硬件资源的使用效率。本文首先分析国内外集群作业调度策略和虚拟化技术的研究现状,并结合用户的实际需求,确定了论文的研究方向和内容。其次研究了OpenLava系统的总体框架、后台守护进程、作业调度原理,通过对用户需求的具体分析,提出了基于OpenLava的虚拟机调度子系统方案,子系统按功能划分为三个模块,分别是虚拟机配置模块、虚拟机克隆模块和虚拟机调度模块,并设计实现了各模块的功能。最后对虚拟机调度子系统进行测试,当集群环境中没有Windows计算节点时,用户向集群提交一个需要Windows运行环境的作业时,虚拟机调度子系统可以动态将Windows虚拟机加入到高性能计算的集群中,并将作业派发到此虚拟机执行,待作业执行完毕后,Windows虚拟机自动关闭,退出集群且释放资源。即实现了针对不同的系统应用提供合适的计算节点资源,解决了用户集群中因缺少Windows计算节点的资源问题,并且通过引入虚拟化技术提高了硬件资源的使用效率。
[Abstract]:With the development of high performance computing, computer cluster system has excellent performance and price ratio, good scalability and high availability. Nowadays, it has become a hot issue of high-performance computer architecture. In the cluster system, the distributed resources in the cluster are managed by the cluster job scheduling system. According to the resource requirement of the job, we can allocate the resources reasonably, make full use of the cluster computing resources, and improve the resource utilization of the system. This paper is based on the OpenLava cluster job scheduling system, combined with the actual needs of users. The virtual machine scheduling subsystem is designed and implemented, which solves the problem that Windows application jobs can not be executed in Linux high performance cluster due to the lack of Windows computing nodes. The purpose of providing appropriate computing node resources for different system application jobs is realized. Firstly, this paper analyzes the research status of cluster job scheduling strategy and virtualization technology at home and abroad, and combines the actual needs of users. This paper determines the research direction and content of the paper. Secondly, the overall framework of OpenLava system, background daemon, job scheduling principle, through the specific analysis of user needs. This paper proposes a scheme of virtual machine scheduling subsystem based on OpenLava. The subsystem is divided into three modules according to function, namely virtual machine configuration module, virtual machine clone module and virtual machine scheduling module. Finally, the virtual machine scheduling subsystem is tested, when there is no Windows computing node in the cluster environment. When the user submits a job to the cluster that needs the Windows environment, the virtual machine scheduling subsystem can dynamically add the Windows virtual machine to the high performance computing cluster. The job is dispatched to this virtual machine to execute, and the Windows virtual machine is automatically shut down when the job is finished. Exit the cluster and release the resources, that is, to provide the appropriate computing node resources for different system applications, and solve the problem of the lack of Windows computing nodes in the user cluster. And through the introduction of virtualization technology to improve the efficiency of the use of hardware resources.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP302

【相似文献】