Heuristics for dynamic topologies to reduce power consumption of networks

Abstract

Energy consumption of communication networks is an important contributor to the ICT sector's greenhouse gas emission footprint. Networks are generally dimensioned for peak loads. Over long periods, networks are under utilised, and at the same time their energy consumption remains high. This research focuses on the reduction of power consumption of communication networks by adapting network topology to traffic demands. Dynamic topologies refer to a method of changing network links and notes according to traffic loads. This paper investigates two heuristics: the Lightest Node First and the Least Loaded Node algorithms that find topologies for given traffic loads, that have a smaller energy footprint, but are able to accommodate traffic loads. Numerical results are presented for a sample network with a large set of traffic demands. Depending on overall network utilisation, the algorithms are able to reduce the average network power consumption by up to 40%.