Skip to Main Content
Cloud computing, a paradigm shift that offer computing resources in a pay-as-you-go manner, is expected to minimize service operators' cost without sacrificing the performance of services. However, it is much harder to guarantee the performance of cloud applications due to the complex architecture of cloud, the interaction between co-deployed applications and the unpredictability of workload. Some research work has reported that existed cloud is not performing as satisfactorily as expected. How to guarantee the stable performance of cloud applications turns into a major concern of both the cloud providers and the service providers. In this paper, we present an approach to guarantee the performance of cloud applications based on a performance model. Extracted from actual runtime monitoring data using data mining techniques, the model can be leveraged to direct how to adjust the resource provision strategy under a given performance requirement. A monitoring framework is implemented for collecting runtime monitoring data on a real cloud which provides PaaS service.
Date of Conference: 18-22 July 2011