Energy Efficient Routing by Switching-Off Network Interfaces

Frédéric Giroire, Dorian Mazauric, Joanna MouliéracAdvances 2013 · DOI · arXiv · OpenAlex

Abstract

Several studies exhibit that the traffic load of routers only has a small influence on their energy consumption. Hence, the power consumption in networks is strongly related to the number of active network elements, such as interfaces, line cards, and base chassis. This paper considers a simplified architecture where a connection between two routers is represented as a link joining two network interfaces. When a connection is not used, both network interfaces can be turned off. The goal is to find the routing that minimizes the number of used links while satisfying all traffic demands. We formally define the Minimum Edges Routing Problem, prove it is not in APX, and propose a heuristic algorithm. Our study on specific topologies and existing backbone networks shows that more than one-third of network interfaces can be spared, leading to significant energy reductions (e.g., 33 MWh for a medium-sized network).

Introduction: The Energy Cost of Connectivity

Information and Communication Technology (ICT) accounts for a significant portion of global energy consumption, with network infrastructure being a major contributor. Empirical studies show that a router's power draw is largely independent of its traffic load—an idle router consumes almost as much power as one at 75% capacity. The real energy "leaks" are the active components: line cards, chassis, and network interfaces.

To address this, we tackle the Minimum Edges Routing Problem. The objective is to route a set of traffic demands using the minimum number of active links (and thus interfaces) without exceeding link capacities. This departs from traditional routing, which prioritizes latency or throughput by spreading traffic across many links. Instead, energy-efficient routing seeks to consolidate traffic onto the fewest possible links, allowing unused interfaces to be powered down.

Energy-efficient routing through traffic consolidation.

Methodology and Complexity Analysis

We formalize the problem as a multicommodity integral flow problem and provide an Integer Linear Program (ILP) formulation for optimal solutions. However, complexity analysis proves that the problem is not only NP-hard but also not in APX. This means that, unless P=NP, there is no polynomial-time algorithm that can guarantee a solution within a constant factor of the optimum.

Complexity proof construction for the Minimum Edges Routing Problem.

To handle large-scale networks, we propose the Less Loaded Edge Heuristic. This algorithm iteratively attempts to remove the least-utilized edges from the network, re-routing traffic each time, until no more edges can be removed without violating capacity constraints. We also derived theoretical bounds for specific graph structures—trees, grids, and complete graphs—to understand the "limit behaviors" of energy-efficient routing.

Key Results and Impact

Evaluation using realistic network topologies from SNDLib (e.g., Atlanta, New York, Cost266) yielded several key insights:

  • Significant Savings: For almost all tested topologies, more than one-third (33%) of network interfaces could be switched off under typical operating conditions.
  • Real-World Impact: In a medium-sized backbone network like Cost266, this approach could save approximately 33 MWh per year.
  • Heuristic Performance: The "Less Loaded Edge Heuristic" performed remarkably well, often reaching within a few percent of the optimal ILP solution while running in milliseconds.
  • Manageable Latency: While paths got longer, the average "stretch" (increase in path length) was only about 27%—a potentially acceptable trade-off for the massive energy gains.

Discussion and Future Work

There is an inherent trade-off between energy efficiency and network resilience. Consolidating traffic increases the stretch factor and reduces fault tolerance by decreasing the number of disjoint paths. Furthermore, while our model assumes a link equals two interfaces, real-world hardware is organized into line cards and chassis. Switching off an entire line card saves significantly more power than switching off a single interface.

Future research will incorporate more complex cost functions reflecting these hardware realities, handle dynamic traffic demands, and move toward experimental validation on physical topologies to measure actual power draw in real-time.

Summary updated

Abstract

Several studies exhibit that the traffic load of routers only has a small influence on their energy consumption. Hence, the power consumption in networks is strongly related to the number of active network elements, such as interfaces, line cards, base chassis, etc. The goal is to find a routing that minimizes the (weighted) number of active network elements used when routing. In this chapter, the authors consider a simplified architecture where a connection between two routers is represented as a link joining two network interfaces. When a connection is not used, both network interfaces can be turned off. Therefore, in order to reduce power consumption, the goal is to find the routing that minimizes the number of used links while satisfying all the demands. The authors first define formally the problem and model it as an integer linear program. Then, they prove that this problem is not in APX, that is there is no polynomial-time constant-factor approximation algorithm. They propose a heuristic algorithm for this problem and also prove some negative results about basic greedy and probabilistic algorithms. Thus, the authors present a study on specific topologies, such as trees, grids, and complete graphs, that provides bounds and results useful for real topologies. They then exhibit the gain in terms of number of network interfaces (leading to a global reduction of approximately 33 MWh for a medium-sized backbone network) for a set of existing network topologies: the authors see that for almost all topologies more than one third of the network interfaces can be spared for usual ranges of operation. Finally, the authors discuss the impact of energy efficient routing on the stretch factor and on fault tolerance.
BibTeX
@inproceedings{giroire2013energy,
  author = {Frédéric Giroire and Dorian Mazauric and Joanna Mouliérac},
  title = {{Energy Efficient Routing by Switching-Off Network Interfaces}},
  booktitle = {Advances},
  year = {2013},
  doi = {10.4018/978-1-4666-1842-8.ch010},
}

Loading related content...

Figure