A High-Grained Traffic Prediction for Microseconds Power Control in Energy-Aware Routers

Recently, as significant increase of Internet traffic, power consumption of ICT devices is growing dramatically. Energy-saving of routers is one of important problems in future networks. There are some studies to reduce the power consumption by adjusting routers' performance according to the volume of incoming/outgoing traffic. In such routers high-grained and accurate prediction of future traffic is very important for controlling power savings. In this paper, we propose a traffic prediction method suitable for performance adjustable routers which achieves accurate traffic prediction in short-term. We discuss about the impacts of prediction parameters and their tuning methods, for accurate of prediction. Through trace-driven simulations with real traffic, we show that our prediction method can reduce up to 95% of power consumption without any packet loss.