What-if QoS Prediction of Cloud-hosted Web Services via Domain Adaptation in Evolutionary Scenarios

摘要

In order to ensure that the web application can continue to provide high-quality services after implementing the bandwidth management schemes, the administrators usually need to predict the Quality of Service(QoS) in advance according to the hypothetical changes of the web system. In the existing research, few pay attention to the prediction of QoS in bandwidth-driven evolutionary scenarios. In this paper, we propose a solution comprised of automated data mining skills and transfer learning techniques to predict the response time of web services in bandwidth-driven evolutionary scenarios. We choose a suitable approach of domain adaptation according to the characteristics of the QoS prediction problems in this paper. There are mainly three contributions in this paper: 1) we adopt an automated data mining approach to extract features for prediction. 2) To our knowledge, it is the first time to apply domain adaptation to the QoS prediction of web services in evolutionary scenarios. 3) We perform some experiments to evaluate the effectiveness and stability of the proposed solution, including two types of evolutionary scenarios under a realworld web application.