A Low-Cost Methodology for Profiling the Power Consumption of Network Equipment

Rate adaptation technologies aim at establishing a linear relationship between power consumption and traffic load in packet networks. They rely on power profiles of network elements, which map system configurations and traffic loads onto power consumption levels, for the selection of network resources to place into low-power states and to identify new system designs with high power-saving yields. We introduce a methodology for profiling the power consumption of network systems that reconciles modeling accuracy with cost containment and rapidity in the preparation and execution of power measurements. We apply the methodology to network systems from multiple vendors and find it capable of delivering a clear message: the power savings enabled by protocol and system software upgrades that support demandtimescale rate adaptation are worthwhile, but also largely inferior to those attainable with a new generation of hardware platforms that pervasively deploy packet-timescale rate adaptation.