Um sistema de Engenharia de Tráfego utilizando Algoritmos Genéticos e Colônias de Formigas

Abstract

The Software Defined Networks (SDN) paradigm is based on the physical separation of frame forwarding control functions. The key idea of ​​separating the programming from the control functions while the specialized hardware for switching frames at high speed remains unchanged. A possible application of this paradigm is in Traffic Engineering (TE). TE aims to optimize the performance of a network, analyzing, predicting and regulating the behavior of transmitted data. This work aims to develop a TE System in SDN fed with data, voice and video traffic sources. Such a system must meet the needs of a network with the characteristics of an Autonomous Internet System. Obtaining the paths for forwarding the traffic of the applications was performed using the Dijkstra algorithm and the heuristics ACO (Ant Colony Optimization) and AG (Genetic Algorithm). The Mininet emulator was used to create the SDN topologies. The use of the solutions obtained by ACO+AG provided a better load balancing and homogeneous occupation of the links. Furthermore, the values ​​of the traffic flows did not exceed the limits of bandwidth of the links. The solutions obtained also meet the maximum end-to-end delay required by the applications.