Energy Saving in IP over WDM Backbone Networks

The energy consumption of the Internet is exploding due to the increase of the number of devices connected to it as well as the increase of traffic volume that is estimated in about 50% per year fuelled by mobile data and video applications; consequently, it is crucial to improve network efficiency to prevent the Internet from being throttled by an energy bottleneck. This PhD dissertation deals with the energy efficiency of telecommunication networks and specifically faces the problem of saving energy in the backbone section of the network. The main contributions concern the definition of algorithms and mechanisms that integrate and exploit the knowledge of the network topology, the carried traffic and the power behaviour of network devices to achieve energy efficiency in the network. Three novel contributions are presented which are related to three different saving strategies. The first faced problem concerns the design of energy efficient virtual topologies in two-layers IP over WDM networks. The problem is formalized as a Mixed Integer Linear Programming (MILP) problem and a novel heuristic algorithm, named Start-Single Hop and ReRoute, is proposed and compared with the optimal solution and another heuristic solution proposed in the literature. Results obtained through extended simulations demonstrate that the proposed solution is able to reduce the network energy consumption with respect to other solutions and to perform close to the optimal solution. Another key aspect faced in the dissertation is related to the possibility of saving energy in presence of variable traffic exploiting traffic engineering strategies that aim at aggregating the traffic on a subset of network elements in order to put in sleeping unused devices. Concerning this point, an Energy Aware Traffic Engineering mechanism, named DAISIES, is proposed. Besides achieving good level of energy efficiency, the proposed solution overcome many practical issues rising from the adaptation of link switch-off mechanism, such as packet loss, out-of-order packet delivering and routing instabilities. The last faced problem concerns the energy efficiency of the optical network layer. In this area a two different solutions are proposed. The first one is based on an iterative greedy algorithm that a posteriori tries to aggregate lightpaths on a subset of optical fibres links, re-optimizing the network when the traffic decreases. The second one, instead, directly works on lightpaths provisioning, exploiting a Power Aware Routing and Wavelength Assignment (PA-RWA) algorithm that tries to set up lightpaths in such a way that the total consumed energy is minimized. Specifically, the performance of the proposed PA-RWA algorithm is evaluated and compared with other solutions proposed in the literature showing better performance both in terms of energy efficiency and blocking probability. In summary, this dissertation makes important contributions to the networking research community providing new methods and approaches for the definition of energy efficient networking techniques.