US DOE Project: Energy Efficiency of Data Networks

Rate adaptation technologies aim at establishing a direct, possibly linear relationship between power consumption and traffic load in packet networks. Since such technologies are typically not deployed in commercial networks as of 2010, it is useful to assess how amenable existing network elements are to their immediate introduction and identify the design upgrades that can maximize their energy savings in new generations of network systems. The formulation of energy profiles for sample commercial equipment is a first step in this direction. In this document we describe how we derive the energy profiles of three commercial network systems, namely one compact Ethernet switch for edge and enterprise applications and two routers for service aggregation in access networks. The profiles are based on a novel formulation of the linear model that defines the overall power consumption of the system as a function of the operating states of all of its components (chassis, line cards, and ports). The results of our experiments indicate that current equipment is not well suited to flow-timescale rate adaptation techniques that remotely control the operating state of parts of a system to minimize the overall power consumption of the network. To create a better match, software and hardware upgrades should be made in system designs to support low-power sleep states for unutilized system components, and to enable packet-timescale rate adaptation methods that instantly adjust the power consumption of data-path hardware to the rate profile of the input traffic.