Um novo algoritmo baseado em aprendizado de máquina para realocação espectral em redes ópticas elásticas multi – núcleos

Abstract

This work proposes a new machine learning (ML)-based algorithm to generate efficient triggers for circuit reallocation in multi-core elastic optical networks. The proposed approach processes network state information and determines the most appropriate moment for optical circuit reallocation, aiming to optimize network performance and minimize the impact of physical layer effects. The results show that the ML model achieves an accuracy of 89% during training. Furthermore, when applying the proposed approach together with another resource management algorithm, an average reduction of 96.6% in circuit blocking probability and 97.3% in bandwidth blocking probability was obtained, with only 0.14% of circuits reallocated per reallocation process.