基于OpenStack的中小企业私有云构建及高可用性研究

发布时间：2018-05-02 09:44

本文选题：云计算 + 私有云　；参考：《华东理工大学》2017年硕士论文

【摘要】：云计算是当今互联网技术的热点之一。利用云计算能够快速改变传统IT基础设施的人工供给方式,实现自动和按需供给,节省了人力资源。同时,利用虚拟化技术一台服务器可虚拟出多个虚拟机供不同用户使用,从而提高了资源的利用率、降低了维护成本。本文通过OpenStack来构建一个私有云平台。针对Openstack默认配置无法应对大规模部署使用的问题,提出一些解决方法,并能显著提高Openstack大部分组件的性能。同时本文也对KVM、XEN以及ESXi虚拟化技术做了各个方面的性能比较,根据综合性能打分,最终选择KVM用于部署私有云系统。高可用是防止核心计算机系统因故障停机的最有效手段之一,Openstack高可用部署方式主要是Active-Passive模式,该模式在生产环境中广泛应用,也是Openstack社区提供的方案,但需要额外维护一台或者多台空闲的备用资源,且当主服务发生故障时,还存在备用服务启动时间较长或启动失败的缺点。本文采用haproxy+ keepalived实现Active-Active模式,该模式下则无需多余的备用资源,所有的资源都是Active状态,发生故障后可以快速切换和迁移,有效提高系统资源利用率。OpenStack默认的调度算法仅仅根据宿主机剩余内存来计算权重,无法满足实际应用中对CPU资源的需求,本文研究了OpenStack的调度算法和执行过程,在调度过滤筛选中增加了Zone筛选条件,并在计算调度权重中增加了变量(CPU使用率),有效提高了OpenStack集群的资源利用率,因此也优化了OpenStack私有云的负载均衡能力。本文选用OpenStack实现企业私有云的部署,对OpenStack的安装部署过程中遇到的问题提供了详细的解决方案,通过性能比较选择KVM作为虚拟化底层软件,深入研究OpenStack高可用部署方案,挖掘OpenStack调度执行流程,并改进OpenStack调度算法,提高OpenStack私有云在生产环境部署的可行性。
[Abstract]:Cloud computing is one of the hotspots of Internet technology. Cloud computing can quickly change the traditional IT infrastructure supply mode, realize automatic and on-demand supply, and save human resources. At the same time, using virtualization technology, a single server can virtual multiple virtual machines for different users, thus improving the utilization of resources and reducing maintenance costs. This article uses OpenStack to build a private cloud platform. Aiming at the problem that Openstack default configuration can not cope with the problem of large-scale deployment, some solutions are put forward, and the performance of most components of Openstack can be improved significantly. At the same time, this paper also compares the performance of KVM Xen and ESXi virtualization technology. According to the comprehensive performance score, we choose KVM to deploy private cloud system. High availability is one of the most effective ways to prevent the core computer system from shutting down due to failure. The high availability deployment mode of Openstack is mainly Active-Passive mode, which is widely used in production environment and is also a solution provided by Openstack community. However, it is necessary to maintain one or more spare spare resources, and when the main service fails, there is the disadvantage of long startup time or failure of the standby service. In this paper, haproxy keepalived is used to implement Active-Active mode. In this mode, there is no extra spare resource, all resources are in Active state, and can be switched and migrated quickly after failure. The default scheduling algorithm of OpenStack can only calculate the weight according to the residual memory of host computer, which can not meet the demand of CPU resource in practical application. This paper studies the scheduling algorithm and execution process of OpenStack. The Zone filter condition is added in the scheduling filter, and the utilization rate of the OpenStack is increased in the calculation of the scheduling weight, which effectively improves the resource utilization of the OpenStack cluster, so it also optimizes the load balancing ability of the OpenStack private cloud. This paper chooses OpenStack to implement the deployment of enterprise private cloud, and provides a detailed solution to the problems encountered in the installation and deployment of OpenStack. Through performance comparison, KVM is chosen as the underlying virtualization software, and the high availability deployment scheme of OpenStack is deeply studied. Mining OpenStack scheduling execution flow and improving OpenStack scheduling algorithm to improve the feasibility of OpenStack private cloud deployment in production environment.
【学位授予单位】：华东理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09;F276.3

【相似文献】