On the Energy Efficiency of Sleeping and Rate Adaptation for Network Devices

Best Paper AwardInternational audienceThe ever-growing appetite of Internet applications for network resources has led to an unprecedented electricity bill for these telecommunication infrastructures. Several techniques have been developed to improve the energy consumption of network devices. As their utilization highly varies over time, the two main techniques for saving energy, namely sleeping and rate adaptation, exploits the lower work-load periods to either put to sleep some hardware elements or adapt the network rate to the actual traffic level. In this paper, we compare two emblematic approaches of these energy-efficient techniques: Low Power Idle and Adaptive Link Rate. Our simulation-based study quantifies the reachable energy savings of these two approaches depending on the traffic characteristics. We show that, with little impact on the Quality of Service and consequent energy savings, Low Power Idle has a clear advantage. On the contrary, ALR is almost always consuming more than LPI and can reach unacceptable QoS levels. We also show that they can be combined to achieve better energy-efficiency, but at the cost of important QoS degradation