Openstack Performance Evaluation System Based On Rally

Cloud computing solves the problems of waste and difficult management of resources in traditional computing models.Cloud computing solutions represented by OpenStack are particularly prominent.But it also brings many problems,such as how to evaluate the performance of OpenStack.OpenStack needs to provide services for a variety of users.Users need to know the performance of the cloud services they have purchased.At the same time,vendors also must quickly know which part needs to be improved in OpenStack.Therefore,the performance evaluation of OpenStack becomes more important,but how to accurately and comprehensively evaluate the performance of OpenStack is a difficult point.At present,the tools for evaluating the performance of OpenStack have these two issues:the first is there is no solution for comprehensive scenarios of Openstack;the second is there is no tools to simulate OpenStack high-load scenarios.Therefore,the following researches are focused on these two issues in this thesis:1)Analyze and design a comprehensive performance evaluation solution based on Rally.Since Rally is a framework of the open source OpenStack performance evaluation,based on Rally we implement plugins for these scenarios for each service component and each object resource of OpenStack,and then form a comprehensive OpenStack performance evaluation solution.We have deployed an evaluation environment,and analyzed the evaluation results of the two OpenStack cloud platforms.2)Design and implemente an automatic,multi-load performance evaluation system.We use a consistent Hash algorithm to implement the stress test cluster for distributed pressure to meet the needs of high-load evaluation scenarios.The system not only performs evaluation on each component of OpenStack,but also performs evaluation based on the combination scenario in between the two components.We have verified that the system could comprehensively and accurately evaluate the performance of OpenStack by functional testing and accuracy testing of performance evaluation.