Traffic concentration for a green Internet

Abstract

Cooling equipment accounts for a sizable fraction of Internet energy footprint. Existing network design models, that assume a linear router power spectrum, are mislead by ignoring the impact of cooling consumption. In this paper, we make a case that the overall energy consumed by a router is polynomial to traffic demand and increases rapidly when the router is loaded. An energy-smart network design is thus formulated to mostly spread out traffic evenly among routers across the network. We further propose a heuristic and compare our design with popular traffic concentration model on two published backbone networks: NSFNET and AT&T. The simulation results reveal that our design saves at least 25% of the consumed energy, and the proposed heuristic closely matches the optimal results. Thus, mitigating network bottleneck routers, rather than creating ones, leads to a greener Internet backbone network. Our conclusions challenge the common wisdom about existing energy saving protocols.