On Modelling the Power Consumption of a Backbone Network

In the last decade, the impact of energy consumption on the environment became more apparent. As the Internet has grown over the past, also it's power consumption has increased. Making the Internet more efficient in terms of power consumption can provide a great contribution in making the world more sustainable. Several approaches have been proposed which save energy at different components and mechanisms. One approach is to improve traffic engineering in order to save energy during times of low traffic demands. To do so, researchers need to rely on a well-suited energy model. By comparing different traffic engineering approaches, we found out that many of them use different algorithmic approaches. But all of them make nearly the same assumptions about the energy consumption of a network. In this paper, we validate these assumptions in the context of modern hardware and propose a network model, that takes the hardware into account. To find a suitable target function, we measured the power consumption of backbone hardware components at different workloads. Furthermore, we measured the impact of distributing the load over different components on the power consumption. With these results, we propose a target function for traffic engineering, that captures the behavior of modern hardware.