支持风险偏好的Web 服务动态组合方法
摘要
With the prevalence of services-oriented architecture, cloud computing and software as a service, there is an increasing number of service providers offering the same or similar functionality but differing in QoS (Quality of Services). Therefore, appropriate service providers should be dynamically selected for the abstract services of a composite service with the objective of optimizing the overall QoS satisfaction. However, traditional service composition approaches select an identical service binding solution for all of the service requests until the next self-adaptation. In addition, these approaches only consider the expected QoS of candidate services but neglect the potential risk from the dynamic QoS fluctuation. As a result, the SLA (Service Level Agreement) between service requesters and providers may be frequently violated, thus affecting the business value of composite service providers. In this paper, we propose a dynamic Web service composition approach that supports different risk appetites. The proposed approach considers both the expected value and fluctuation of QoS, and applies portfolio theory to generate multiple service binding solutions satisfying the given risk appetite, thereby controlling the risk and adapting to different risk appetites. Compared with global service selection approaches, the experimental study shows that our approach can effectively reduce the SLA violations and improve the business value of composite services.