End-to-End Data Flows Management in the Decentralized 5G/6G Mobile Networks

Abstract

The evolution of 5G and the anticipated emergence of 6G networks demand significant enhancements in optical backhaul infrastructure to support higher bandwidth, low latency, and increased reliability. Considering a wide range of feasible applications with different quality requirements, the key challenge for the underlying backhaul and optical transport network is in the high load variance in the optical switching nodes and complicated data flows management. In this paper, a novel approach is proposed for end-to-end data flows management in decentralized 5G/6G mobile networks, which are interconnected by the optical burst switching transport infrastructure. The key idea is to train a deep recurrent neural network over a real network statistic obtain at the different network segments and service slices. Then, predictions made by deep neural network are used for predictive resource allocation in each node of the optical burst switching network to ensure a target quality of service for each end-to-end data flow. The experimental results show that proposed approach provides 90% accuracy of predictions and allows to effectively utilize the resources of optical network.