Minimizing power consumption in SDNs: measurements and optimization

Large-scale IT infrastructures are highly energy-intensive systems. To mitigate the environmental impact of networks, it is crucial to design energy-aware traffic engineering strategies that make the best use of a network's redundancy in order to maximize energy savings. The Software-Defined Networking (SDN) paradigm is a powerful tool that eases network management by decoupling the control and data planes. Because of its flexibility and controllability, SDN is nowadays widely adopted in data centers and enterprise networks. In this paper, we provide energy-aware traffic solutions for SDNs. We conducted extensive experiments on real switches for a thorough power consumption characterisation. Thanks to this preliminary study, we could fully characterize the solution space of a performance-constrained energy optimization problem. We formulate two optimization problems for the selective activation of switches and ports under hard constraints on traffic demand, considering both a static and dynamic traffic scenario. We show that the proposed problems are NP-hard and provide two polynomial-time heuristics for the static and dynamic case, respectively. Through simulations, we show that our solution outperforms previous approaches in all the considered settings.