ICTON 2011 Mo.C3.1 978-1-4577-0882-4/11/$26.00 ©2011 IEEE 1 Advantages of Using Cognition when Solving Impairment-Aware Virtual Topology Design Problems R.J. Durán 1 , N. Fernández 1 , I. de Miguel 1 , M. Angelou 2 , D. Sánchez 3 , J.C. Aguado 1 , T. Jiménez 1 , P. Fernández 1 , N. Merayo 1 , N. Atallah 3 , R.M. Lorenzo 3 , I. Tomkos 2 , E.J. Abril 1 1 Optical Communications Group, University of Valladolid, Paseo de Belén 15, 47011 Valladolid, Spain Tel: +34 983 423000 ext. 5557, Fax: +34 983 423667, e-mail: rduran@tel.uva.es 2 AIT-Athens Information Technology centre, Tel: +302106682773, e-mail: itom@ait.edu.gr 3 CEDETEL-Center for the Development of Telecommunications, Tel: +34983546502, e-mail: ruben.lorenzo@cedetel.es ABSTRACT In this paper, the advantages of using cognition when solving the impairment-aware virtual topology design problem are demonstrated. To this end, an algorithm to design the virtual topology, GAPDELT, previously proposed by the University of Valladolid, has been extended to ensure that all the lightpaths of the virtual topology comply with quality of transmission constraints. The new version, called IA-GAPDELT, is a multiobjective genetic algorithm which uses Pareto optimality to reduce both the network congestion and the number of transmitters in operation, and so the energy consumption. By means of simulation, we show that when the algorithm is enhanced with a simple cognitive technique, it obtains a higher number of feasible solutions (virtual topologies) and, moreover, they are generally better in terms of the optimization parameters, than those obtained without cognition. Keywords: Virtual topology design, impairment aware, cognitive process, genetic algorithm, Pareto optimality. 1. INTRODUCTION The high dynamism and flexibility that future optical networks should have makes that the virtual (or logical) topology (i.e., the set of lightpaths established in the network [1]) should fulfil two conditions. Firstly, it has to be efficient in terms of capacity, energy consumption and quality of transmission of the connections established. Secondly, it should be adapted to traffic and network conditions. Thus, it is necessary to develop methods to design efficient virtual topologies in a relatively short period of time, so that reconfiguration mechanisms can be triggered on real time if required. Moreover, the virtual topology design problem is a multiobjective optimization problem, since several parameters should be optimized rather a single one. As there are trade-offs between different solutions in terms of the different parameters, the most interesting approaches to solve multiobjective optimization problems are targeted to obtaining the Pareto optimal set, i.e., a set of solutions where each solution (e.g., a virtual topology) is characterized because it cannot be simultaneously improved in terms of all the optimization objectives. In [2], we presented GAPDELT (Genetic Algorithm to Provision the network and to DEsign the Logical Topology), which allocates the transmitters and receivers required in each network node, determines the number of wavelengths needed per link, and designs the virtual topology (determining which nodes should be connected by means of lightpaths, finding a route and a wavelength for each lightpath, and routing the traffic through those lightpaths). The objectives of GAPDELT are to minimise both the congestion, i.e., the traffic carried by the most loaded lightpath, and the number of resources required, and it provides as a result a collection of virtual topologies which constitute a good estimate of the Pareto optimal set. We have now improved that work by developing a new version of that algorithm, called IA-GAPDELT (Impairment Aware – GAPDELT), which ensures that the Q-factors [3] of all the lightpaths of the virtual topology are higher than a user-defined threshold. IA-GAPDELT is again a multiobjective genetic algorithm to design virtual topologies minimizing the congestion and the number of transmitters in operation. In this way, IA-GAPDELT optimizes the network capacity (by minimizing congestion) and reduces the energy consumption (by reducing the number of transmitters in operation) while ensuring that all the lightpaths established fulfil the quality of transmission constraint on the Q-factor. The objective of this paper is to demonstrate the advantage of using cognition in the methods to design the virtual topology, mainly when the joint optimization of several parameters is required and when the algorithm should not take too much time to obtain the solutions in order to be used in reconfigurable Wavelength-Routed Optical Networks (WRONs). Hence, a new version of IA-GAPDELT is also presented to this end. It is called SC-IA-GAPDELT (Simple Cognition – IA-GAPDELT), as it uses a simple cognitive technique to improve the performance of the algorithm. By means of simulation results, the positive impact of cognition is demonstrated, as SC-IA-GAPDELT obtains a better estimate of the Pareto optimal set: not only a higher number of virtual topologies of the set are obtained, but they are generally better (in terms of the optimization parameters) than those obtained in the version without cognition.