Conserving Power Consumption in Elastic Optical Networks Using Deep Learning

The power consumption issue in elastic optical networks is a prominent topic of widespread attention and concern nowadays. The wasteful on-and-off transitions of the networking components including transponders, optical cross-connects, and amplifiers are one of the major power consumers in elastic optical networks. Turn-on transitions may result in power consumption spikes that are more than 4 times the needed amount when they are active. Most currently employed power control strategies are not designed to handle this significant power consumption. To solve this problem, in this paper, the number of active lightpaths crossing an element within a short period of time is predicted using the long short-term memory technique. This knowledge is used to avoid the frequent deactivation and activation of the components. Using numerical simulations, we demonstrate that our proposed scheme substantially improves the average power consumption in NSFNET and USNET topologies.