Algoritmo genético para o roteamento de tráfegos elefante em redes de data center

Abstract

As computer networks increase in size, new challenges in routing data arise. It becomes important to employ strategies that seek to reduce data congestion and packet loss. To provide an alternative to existing routing algorithms, this paper proposes a genetic algorithm to route elephant flows in data center networks. This algorithm seeks to minimize the utilization of links in networks with fat-tree topology in an attempt to perform load balancing. It was implemented in a controller for SDN architecture networks with emulated virtual networks. The algorithm was tested with different parameters for a set of traffic patterns. The best results, according to the chosen metrics, were compared to other related routing algorithms. The comparisons showed that the proposed algorithm was able to improve the efficiency of the tested networks for certain traffic patterns.