Research On Theautomatic Deployment And System Management Technology For Large-scale Openstack Cluster

In the last few years, cloud computing has become the most pop research fieldboth in industrial and academic circles. The rapid development of cloud computinghas promoted the changes of traditional IT infrastructure. Various IT enterprises,scientific research institutions and telecom service providers with large-scale serverrequirements pay more and more attention to their construction of the cloudcomputing infrastructure. As an open source cloud computing management platform,OpenStack has been widely deployed and applied in all walks of life due to itsfeatures of user cost saving, high operational efficiency, enterprise innovation benefit,etc. However, the deployment and management of a large-scale OpenStack cluster isa time-consuming and error-prone task, which is mainly caused by the complexdependencies and abundant configuration parameters of the OpenStack componentsas well as the large-scale of the cluster. Therefore, how to deploy the large-scaleOpenStack cluster quickly and automatically has become a topic which is worthy ofin-depth study.Based on the in-depth research of the OpenStack architecture and theory, thisthesis contrasts and analyzes the automatic deployment and management technologyof traditional cluster and the existing OpenStack cluster. This thesis designs andimplements an automatic deployment system of the large-scale OpenStack clusterwhich is named MMADS on the basis of the SaltStack configuration managementframework. This system aims to automatically deploy OpenStack relatedcomponents according to the user’s personalized configuration parameters and thedescription of the cluster nodes’ roles. The advanced natures of this thesis mainlydisplay in the following four aspects:(1) MMADS system provides a lightweight configuration interface which canprovide the intuitive and convenient cluster configuration method for users andexternal applications.(2) MMADS system provides an automatic deployment method based on theroles of the cluster nodes. Users can specify the roles of cluster nodes statically ordynamically with the cluster deployment algorithm.(3) MMADS system adopts SaltStack configuration management framework to manage the state of the cluster nodes which can effectively manage the underlyingresources’life cycle and dependencies.(4) MMADS system provides the extension mechanism of multiple Masterservice which can effectively improve the efficiency of deployment for thelarge-scale OpenStack cluster.