Improving Cloud Service Reliability -- A System Accounting Approach

Nowadays an increasing number of companies only deploy their enterprise application services over the Internet. Software as a Service (SaaS) in a cloud computing environment allows these companies to focus on providing more competitive services instead of maintenance. As delivery of computing as a service, a trustworthy cloud service widely depends upon its reliability. For this reason, a newly defined Quality of Reliability (QoR) for cloud services is proposed in this paper. To achieve a good QoR, we not only analyze system events from both service consumers and providers, but also provide a layered composable system accounting architecture for cloud systems. A pipelined approach and a dependence estimation algorithm are introduced for pattern recognition and event analysis and prediction. A self-healing layer is also designed to achieve automatic recovery by re-composing services according to their functionalities and non-functional requirements. An implementation of this framework in an education services environment confirms the advantages over extant system accounting systems.