Investigation of an automatic deployment transformation method for OpenStack

Abstract: Cloud computing is the on-demand availability of computer resources provided as a service over a network. OpenStack is an open-source cloud computing software. Deploying and operating OpenStack manually is a tedious process. To address this,life-cycle management tools have been developed. These tools automate the process of deploying OpenStack and can work as operations and maintenance tools. As OpenStack follows a six-month release cycle, some of the life-cycle management tools can not keep up with the releases and end up outdated due to a lack of support from the OpenStack community. This leads to older OpenStack deployments being stuck on unsupported life-cycle management tools, which could have bugs, security issues and are often more complicated to manage than newer life-cycle management tools(LCMTs). One way to solve this is by moving the OpenStack deployment from one LCMT to another, that is migration of the deployment itself. This thesis addresses the issue by identifying the current popular LCMTs through a secondary survey by OpenStack foundation and the existing migration methods through literature review. Furthermore, the effect of LCMTs on the OpenStack deployment is analysed, and controlled experiments are performed to test non-live migration between different LCMTs based OpenStack deployments. The results from the OpenStack user survey shows that, Kolla-ansible, followed by Puppet and OpenStack-ansible are the current popular LCMTs, based on their usage amongst the survey participants. The literature review combined with experimentation shows that the existing migration models are limited to the LCMT environments and the LCMTs themselves effect the OpenStack deployment in deployment file locations and through underlying technologies. We also propose an experimental method which works for migrating OpenStack from OpenStack-Ansible to Kolla-Ansible through a Manual deployment and vice-versa, which can thereby be generalized.