Machine Learning Assisted Traffic-Aware Approach to Path Assignment in SDM-EONs

Abstract

The introduction of new technologies and applications connected to the Internet has demonstrated the inability of current optical networks to provide resources for next-generation Internet. Although the emergence of elastic optical networks with space-division multiplexing has shown to be a promising solution to deal with the capacity problem, some of the technical requirements for the implementation of these networks remain open challenges. In this sense, this paper proposes MISSION, a Machine Learning assisted, fragmentation, and crosstalk-aware model for path allocation in Space Division Multiplexing Elastic Optical Networks (SDM-EONs). The proposed approach is capable of ordering candidate paths for allocation based on metrics such as crosstalk, fragmentation, and the number of slots. Besides, MISSION shows competitive performance, by keeping a comparatively low blocking probability and fragmentation, even under heavy loads.