Otimização de rotas em redes definidas por software utilizando algoritmos evolucionários

Abstract

The popularization of cloud services and applications has drastically changed Internet traffic profiles. Such changes have motivated the evolution of the SDN (Software Defined Networking) paradigm, where the data and control planes are separated from the switching elements of the networks, enabling a centralized management view and a series of new applications. From this perspective, optimization models were developed in this work, based on evolutionary algorithms, for routing in best effort and with SLA (Service Level Agreement) scenarios, meeting load balancing and energy efficiency criteria in the use of resources. Optimization models were developed and evaluated in different contexts, considering the quality of the achieved solutions and the performance of the methods. After several analyses, the NSGA-II (Non-dominated Sorting Genetic Algorithm II) was chosen for use in obtaining sets of optimal solutions. In the case of SLA, the objective aggregation technique with a genetic algorithm was chosen, defining a priori preferences. Experiments performed through emulation and simulation indicated an improvement in the performance of the networks, according to the preferences defined by a decision maker, with the algorithm showing the ability to converge to routes that meet the restrictions of the network’s flow demands, also ensuring a minimum use of resources, aiming at energy efficiency.