Ant colony based self-adaptive energy saving routing for energy efficient Internet Young-Min Kim a,⇑ , Eun-Jung Lee b , Hea-Sook Park a , Jun-Kyun Choi b , Hong-Shik Park b a Electronics and Telecommunications Research Institute (ETRI), Republic of Korea b Electrical Engineering Department, Korea Advanced Institute of Science and Technology (KAIST), Republic of Korea article info Article history: Available online 12 April 2012 Keywords: Ant colony optimization Energy saving routing Traffic centrality Pheromone trails abstract According to recent research, the current Internet wastes energy due to an un-optimized network design, which does not consider the energy consumption of network elements such as routers and switches. Looking toward energy saving networks, a generalized prob- lem called the energy consumption minimized network (EMN) had been proposed. How- ever, due to the NP-completeness of this problem, it requires a considerable amount of time to obtain the solution, making it practically intractable for large-scale networks. In this paper, we re-formulate the NP-complete EMN problem into a simpler one using a newly defined concept called ‘traffic centrality’. We then propose a new ant colony-based self-adaptive energy saving routing scheme, referred to as A-ESR, which exploits the ant colony optimization (ACO) method to make the Internet more energy efficient. The pro- posed A-ESR algorithm heuristically solves the re-formulated problem without any super- vised control by allowing the incoming flows to be autonomously aggregated on specific heavily-loaded links and switching off the other lightly-loaded links. Additionally, the A- ESR algorithm adjusts the energy consumption by tuning the aggregation parameter b, which can dramatically reduce the energy consumption during nighttime hours (at the expense of tolerable network delay performance). Another promising capability of this algorithm is that it provides a high degree of self-organizing capabilities due to the amaz- ing advantages of the swarm intelligence of artificial ants. The simulation results in real IP networks show that the proposed A-ESR algorithm performs better than previous algo- rithms in terms of its energy efficiency. The results also show that this efficiency can be adjusted by tuning b. Ó 2012 Elsevier B.V. All rights reserved. 1. Introduction Energy conservation is becoming a key issue for the Internet because its energy consumption is remarkably high and may exponentially grow [1,2]. According to sev- eral studies [3], the global amount of energy consumed by the Internet has been approximately 900 billion kWh, which represents 5.5% of the total global electricity consumption; this consumption is estimated to increase by around 20–25% annually. Furthermore, the energy effi- ciency of the current Internet is very low, approximately 8–10 times lower than that of wireless networks [4], which arises from the fact that, as indicated in [5], the network elements, such as routers and switches, in the current Internet are not optimized for energy saving; they tend to consume the maximum energy even though the carried traffic occupies only a small portion of their capacity, and their capacity is usually over-provisioned for accommodat- ing future growth. Experiments in [6] found this energy inefficiency in current network elements, where is little or no difference in the energy consumption between peak and off-peak periods. 1389-1286/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.comnet.2012.03.024 ⇑ Corresponding author. Address: ETRI, 218 Gajeong-ro, Yuseong-gu, Daejeon 305-700, Republic of Korea. Tel.: +82 42 860 5883, mobile: +82 10 6680 3059; fax: +82 42 860 6342. E-mail addresses: injesus01@kaist.ac.kr, injesus@etri.re.kr (Y.-M. Kim), freakone@kaist.ac.kr (E.-J. Lee), parkhs@etri.re.kr (H.-S. Park), jkchoi@ kaist.ac.kr (J.-K. Choi), parkhs@kaist.ac.kr (H.-S. Park). Computer Networks 56 (2012) 2343–2354 Contents lists available at SciVerse ScienceDirect Computer Networks journal homepage: www.elsevier.com/locate/comnet