Rate adaptation for energy efficiency in packet networks

Abstract

Data networking services and applications enable substantial energy savings in broad sectors of economic activity by substituting the transportation of people and goods with the electronic transfer of data packets. To maximize the efficiency of packet networks, energy use in network equipment should scale rigorously with network traffic load. The two main frameworks for enforcing energy proportionality between energy consumption and network load are rate scaling and sleep state exploitation. Unfortunately, existing instantiations of these methods fail to scale energy consumption over the full load spectrum and require either tight coordination between adjacent nodes or new end-to-end traffic management mechanisms. In this paper, we define a novel approach for combining rate scaling and sleep state exploitation into a single rate adaptation scheme that preserves the best properties of its components in every portion of the load spectrum. We also demonstrate how rate adaptation can be applied to individual nodes by utilizing a simple data path architecture and show that rate adaptation is a feasible mechanism for achieving energy proportionality even when applied to individual systems or subsystems.