Towards Source Selection in Transfer Learning for Cloud Performance Prediction

Learning performance models for cloud services is challenging due to the dynamics of the operational environment stemming from scaling and migration decisions. This requires exchange or adaptation of the models in order to maintain prediction accuracy over time. Approaches that incorporates previously acquired knowledge using transfer learning is a viable technique for timely and robust model adaptation, especially when the training data is limited. In this paper we quantity the impact of different source domains on the accuracy of a target model in another domain. The evaluation is performed using data traces collected from a testbed that runs a Video-on-Demand service and a Key-Value Store under various load conditions. We find that the choice of source domain can yield a transfer gain, and sometimes a substantial transfer penalty.