Deep Learning based Traffic Optimization in Optical Transport Networks

Rapid technological development and continuous data growth tends to increase the pressure on the existing network infrastructure. In order to tackle this challenge optical transport networks are continuously improved to enhance their capabilities. In this paper, we propose a novel deep learning- based data flows optimization in optical label switched networks. The key idea of the proposed approach is to use data acquisition form the network nodes in order to collect statistical information of the network performance. Statistical information is than used to train deep neural network and determine optimal nodes configuration to improve the overall network performance. Simulations results show that proposed data flows aggregation algorithm allows to improve bandwidth utilization, while providing acceptable latency and packet loss rate.