Link power coordination for energy conservation in complex communication networks

Communication networks consume huge, and rapidly growing, amount of energy. However, a lot of the energy consumption is wasted due to the lack of global link power coordination in these complex systems. This paper proposes several link power coordination schemes to achieve energy-efficient routing by progressively putting some links into energy saving mode and hence aggregating traffic during periods of low traffic load. We show that the achievable energy savings not only depend on the link power coordination schemes, but also on the network topologies. In the random network, there is no scheme that can significantly outperform others. In the scale-free network, when the largest betweenness first (LBF) scheme is used, phase transition of the networks’ transmission capacities during the traffic cooling down phase is observed. Motivated by this, a hybrid link power coordination scheme is proposed to significantly reduce the energy consumption in the scale-free network. In a real Internet Service Provider (ISP)’s router-level Internet topology, however, the smallest betweenness first (SBF) scheme significantly outperforms other schemes.

💡 Research Summary

The paper addresses the growing concern that modern communication networks consume an ever‑increasing amount of electricity, much of which is wasted because links are kept powered even when traffic is low. The authors propose a set of link‑power coordination schemes that progressively place selected links into an energy‑saving mode, thereby concentrating traffic onto a smaller set of active links during off‑peak periods. By doing so, overall network power consumption can be reduced without sacrificing connectivity or throughput.

Four coordination strategies are examined. The first, a Random scheme, deactivates links in an arbitrary order, serving as a baseline. The second, Largest Betweenness First (LBF), turns off the links with the highest betweenness centrality first; these are typically the “highways” of traffic. The third, Smallest Betweenness First (SBF), deactivates the least‑central links first, preserving the backbone while shutting down peripheral edges. Finally, a Hybrid scheme dynamically switches between LBF and SBF based on the current load, aiming to combine the advantages of both.

The authors test these schemes on three representative topologies: (1) an Erdős‑Rényi random graph, (2) a scale‑free network generated by the Barabási‑Albert model, and (3) a real‑world ISP router‑level topology obtained from public measurement data. In the random graph, all schemes perform similarly, achieving modest energy savings (≈10–15 %) because the uniform connectivity provides many alternative paths regardless of which links are turned off. In the scale‑free network, the LBF scheme initially yields large savings during the cooling‑down phase, but as traffic rises it triggers a sharp decline in the network’s transmission capacity—a phase transition caused by the removal of hub links that carry the majority of flow. To mitigate this, the Hybrid scheme first applies SBF to deactivate low‑centrality edges, then switches to LBF only when the load exceeds a predefined threshold. This adaptive approach raises average energy savings to over 30 % while keeping the capacity collapse under control.

In the ISP topology, which exhibits a hierarchical, highly asymmetric structure, the SBF scheme outperforms all others, delivering roughly 28 % power reduction. The backbone links in this network have very high betweenness values; protecting them ensures that service quality remains stable while peripheral links are safely powered down. Moreover, the time required to reactivate a link is a critical operational factor in ISP environments; the gradual, low‑impact deactivation pattern of SBF aligns well with such constraints.

Quantitative results show that the effectiveness of link‑power coordination is strongly topology‑dependent. Random networks offer little room for optimization beyond baseline savings. Scale‑free networks benefit from adaptive strategies that respect the centrality distribution of links, whereas real‑world ISP topologies favor conservative, backbone‑preserving policies. The paper concludes with several practical implications: (i) link‑power policies must be tailored to the structural characteristics of the network; (ii) in heterogeneous networks, anticipating phase transitions and employing hybrid, load‑aware mechanisms can achieve both energy efficiency and robustness; (iii) for operational carrier networks, SBF‑type schemes combined with fast link‑wake‑up mechanisms provide the most realistic path to substantial power reduction without compromising quality of service. Future work is suggested in the areas of real‑time traffic prediction, detailed modeling of link activation/deactivation costs, and multi‑objective optimization that incorporates QoS constraints alongside energy objectives.