DAFNES: A distributed algorithm for network energy saving based on stress-centrality

We focus on the problem of reducing power consumption in backbone networks by putting in sleep mode IP links in order to save energy. We propose a new algorithm, called DAFNES, which relies on the stress centrality index in order to take the switch off decision. Differently from previous work in the literature, our algorithm requires neither the complete knowledge of the traffic matrix nor a careful tuning of the input parameters. Results, obtained over two realistic case studies, prove the efficiency and efficacy of our solution, with more than 50% of power saving while preserving Quality of Service constraints in terms of network connectivity, link congestion avoidance and increase of path lengths. Moreover, we show that the extra overhead required for running the distributed solution is limited compared to the amount of traffic exchanged in the network by the users. Finally, we face different implementation issues, including: (i) the reduction of the number of times our solution is applied, (ii) the evaluation of the algorithm performance on an emulated testbed.