Practical knowledge acquisition on robust IP traffic control

In this paper, we focus on a practical knowledge acquisition centric approach, especially concerning IP traffic engineering (TE) over MPLS (Multi-Protocol Label Switching). Although MPLS has been standardized to solve the network-wide congestion problem, the realization of automated control capabilities is an important problem from the viewpoint of practical use. As for services, the overlay network in application layer is often used. Due to the respective optimal strategy of packet transmission, securing the stability of network behavior becomes a problem more and more. As a main requisite for the solution, the robustness that keeps operating normally in appropriate time for unpredictable behavior on the network is needed in spite of disturbances by the effects of multiple service overlay networks. We analyze the results of practical experiments from the control-theoretic viewpoint based on acquired knowledge regarding oscillating behaviors and the parameter tuning for stabilization., In addition, a simulation with a plant model, an improvement of stabilization using the proportional-integral control, and a feedback compensation with the Smith predictor are described.