Evaluation of Traffic Engineering Based on Model Predictive Control Using Traffic Trace in Actual Network

Traffic engineering with traffic prediction is one approach to accommodate time-varying traffic stably. In this approach, the routes are calculated so as to avoid congestion based on the predicted traffic. However, predicted traffic includes prediction errors. We proposed a traffic engineering method with traffic prediction, which is robust to prediction errors. To achieve the robust control against the prediction errors, our method uses the idea of the Model Predictive Control(MPC), which is a method of process control with predicting the system dynamics. In each control cycle, our method calculates the routes with the new predicted traffic which is corrected by the feedback from the observation. In addition, our method avoids the large routing change to achieve the stable routing even if the prediction error may happen. In this paper, through simulation with the actual traffic trace, we clarify that our method avoids the congestion even if the predicted traffic includes errors.