On the Interplay Between Sleep Modes and Device Lifetime in Telecommunication Networks Luca Chiaraviglio and Marco Listanti DIET at University of Roma “Sapienza”, Rome, Italy. [email protected],[email protected]

Abstract. We study the impact of sleep modes on the device lifetime for backbone, cellular networks and fixed access networks. The device lifetime is defined as the amount of time between one hardware failure and the following one. We show that sleep modes bring to two complementary effects on the device lifetime. Thus, we argue that a more comprehensive approach targeting the device lifetime is a mandatory step towards a sustainable green networking. Keywords: failure rate, device lifetime, energy-aware networks

Energy-efficiency in telecommunication networks has been deeply investigated in the last years. In particular, one of the most promising solutions to save energy is the adoption of sleep modes. The basic idea of this approach is to completely shut down the device during low traffic periods, i.e., at night or during weekends, since the network is normally dimensioned for the peak traffic. The effectiveness of algorithms based on sleep modes has been proved for both backbone networks [Idzikowki - 2013], cellular networks [Ajmone Marsan - 2013] and fixed access networks [Dixit - 2013]. However, none of the previous solutions has taken into account the impact of sleep modes on the lifetime of the device. In this work, we define the device lifetime as the amount of time between one failure and another one. Failures are critical events occurring in telecommunication networks, especially when sleep modes are applied. In such a case, in fact, the overprovisioning of the resources available is reduced, since many networks elements (normally linecards in backbone networks, Base Stations in cellular networks and optical network units) are put into a low power state. Thus, a failure may bring the network to an unacceptable degradation in the Quality of Service, since the amount of resources available for backup routes is limited. In this work, we argue that there is an interplay between sleep mode and the device lifetime in a telecommunication network. We base our claim on two assumptions: i) the failure rate is related to the temperature of the device, ii) the frequency of the change of state (from sleep mode to full power and vice-versa) impacts the failure rate. Both these effects have been studied for memories and CPUs. In particular, a first-order model that predicts the impact of the temEa perature on the failure rate is the Arrhenius law [Laidler - 1977]: γ T = γ 0 e− KT where γ 0 is the failure rate assuming infinite temperature,1 Ea is the activation 1

This term can be estimated by measuring the failure rate at very high temperatures

energy, K is the Boltzmann constant, T is the measured temperature, and γ T is the predicted failure rate. Although current measurements have shown that this model is an approximation of the real behavior [De Salvo - 1999], it has been proven that the temperature increase negatively impacts the failure rate. In our case, when sleep modes are applied, the temperature of the device is reduced: this is due to the fact that most of the components of the device (either for a linecard, Base Station or an optical network unit) can be put in a low-power and consequently in a low-temperature state. However, the frequency of the change of state negatively impacts the failure rate. This effect is known in the literature as thermal cycling: N f = C0 (∆T − ∆T0 )−q where ∆T is temperature variation, ∆T0 is variation of temperature supported by the device, C0 is a constant that depends on the device material, q is the Coffin-Mason exponent, and N f is the number of cycles to failure. The failure rate due to thermal cycling is TC defined as γ ∆T = fN f where f T C is the thermal cycling frequency and γ∆T is the resulting failure rate. Thus, an increase in the frequency brings an increase in the failure rate. In our scenario, frequent transitions between full power and sleep mode will result in an increase in the failure rate. In [Chiaraviglio - 2013] we have performed a first evaluation of the impact of sleep modes for the failure rate in backbone networks. In particular, we have evaluated the failure rate by applying energy-aware algorithms. However, in this work we claim that the minimization of power may deteriorate the failure rate, i.e., frequent transition changes increase the failure rate. Thus, we argue that a sustainable approach should integrate the failure rate in the objective function of the problem. Additionally, we claim that the impact of sleep modes on the failure rate should be investigated also for cellular networks, since these networks normally consume a huge amount of power for the operator. Finally, the interplay between sleep modes and failures should be evaluated also in fixed access networks. As future work, we plan to investigate which components are the most critical ones on the device lifetime when sleep modes are applied. Additionally, we will characterize the failure rate events (hw and sw failures). Finally, we will develop models that integrate the failure rate at the network level.

References [Laidler - 1977] Laidler, K. J.: Chemical kinetics. McGraw-Hill, 1977. [De Salvo - 1999] De Salvo, B., Ghibaudo, G., Pananakakis, G., Reimbold, G., Mondond, F., Guillaumot, B., Candelier, P.: Experimental and theoretical investigation of nonvolatile memory data-retention. IEEE Transactions on Electron Devices, 46(7), 1518-1524, July 1999. [Chiaraviglio - 2013] Chiaraviglio, L., Cianfrani, A., Coiro, A., Listanti, M., Lorincz, J., Polverini, M.: Increasing device lifetime in backbone networks with sleep modes. 21st International Conference on Software, Telecommunications and Computer Networks (SoftCOM), pp. 1-6, September 2013. [Idzikowki - 2013] Idzikowski, F., Bonetto, E., Chiaraviglio, L., Cianfrani, A., Coiro, A., Duque, R., Jimenez, F., Le Rouzic, E., Musumeci, F., Van Heddeghem, W. Lopez Vizcaino, J. Ye, Y.: TREND in Energy-Aware Adaptive Routing Solutions. IEEE Communications Magazine, vol. 51, no. 11, pp. 94-104, November 2013.

[Ajmone Marsan - 2013] Ajmone Marsan, M., Chiaraviglio, L., Ciullo, D., Meo, M.: On the effectiveness of single and multiple base station sleep modes in cellular networks. Computer Networks, 57(17), 3276-3290, December 2013. [Dixit - 2013] A. Dixit, S. Lambert, B. Lannoo, D. Colle, M. Pickavet, P. Demeester: Towards energy efficiency in optical access networks. IEEE ANTS 2013, 7th International Symposium on Advanced Networks and Telecommunication Systems, Chennai, India, December 2013.