Energy-Aware Server Provisioning in Large Scale Video-On-Demand Systems

Video-on-demand has been emerging as a very popular internet service in recent years. But energy consumption is becoming a critical issue as these services scale up. In this paper, we propose an energy-aware server provisioning strategy which dynamically turns on/off servers in order to adaptively tailor active servers to dynamic user load. We initiate a stochastic model which characterizes unique properties such as bandwidth and power consumption of video-on-demand systems. We then employ a measurement-based adaptive online user load predictor and apply large deviation theory to our model to develop global strategy. Simulation confirms that our strategy can lead a significant amount of energy savings with little or no user experience degradation.