Evolution of resource management in elastic optical networks: from RSA and RMSA to AI-driven autonomous networking

Abstract

The global bandwidth requirements are increasing rapidly due to significant growth in cloud computing, next-generation high-capacity communication systems, Internet of Things (IoT) applications, and video streaming services. Initially, the fixed-grid channel allocation mechanisms were sufficient for a limited number of users, but with increasing demand for global bandwidth, the conventional wavelength division multiplexing (WDM)-based optical networks face challenges in efficiently utilizing spectrum resources. To address these limitations, elastic optical networks (EONs) have been developed as a solution for accommodating increasing traffic demands with improved spectral efficiency and network utilization. However, EONs face resource management challenges, including efficient spectrum utilization, complexity in spectrum allocation, and spectrum fragmentation under dynamic traffic conditions. To address these challenges, researchers have proposed routing and spectrum allocation (RSA) techniques that subsequently evolved into routing, modulation, and spectrum allocation (RMSA) strategies that jointly consider routing, modulation, and spectrum allocation. In recent years, research has evolved from RSA/RMSA to machine learning, deep learning, and further AI-based resource management approaches. The paper compares conventional, heuristic, metaheuristic, and AI-based resource management approaches in terms of spectrum utilization, blocking probability, computational complexity, adaptability, and scalability. It provides researchers and network designers with a comprehensive understanding of the evolution of EON resource management, while identifying key research gaps and challenges in the transition toward AI-driven autonomous networking.