Power Aware PoP Design and Auto Configuration

We consider the problem of the minimizing power consumption of Internet backbone networks by the means of power aware design and auto configuration of the Points of Presence (PoPs). We first show that the problem of achieving a power-optimal PoP design under general traffic demands is NP-Hard. The problem is NP-Hard even when the traffic demands are correlated i.e. traffic demands across different source-destination pairs scale uniformly. Under a simplified model that ignores power consumption of PoP-internal links i.e., links between core routers inside a PoP, we propose the Port-Sorting algorithm which outputs a provably optimal PoP design. Further, we propose an auto-configuration scheme, that activates minimal number of network components for a given traffic load. For the case where the power consumption of PoP-internal links is taken into account, we propose a heuristic based on the Port-Sorting algorithm that achieves power-efficient PoP design and auto-configuration. We conduct extensive simulations to evaluate the energy savings due to our algorithms and compare our algorithms to other candidate algorithms. We find that our PoP design and auto-configuration algorithms achieve near-optimal performance over a wide range of traffic instances and significantly outperform other candidate algorithms. Furthermore, when compared to the state-of-the-art, which leaves all network components running at all times, our algorithms demonstrate over 3x in energy savings for typical traffic loads.