Application of Machine Learning in Routing and Wavelength Assignment: Literature Survey

This paper presents the survey of the prominent issue in optical networks is Routing and Wavelength Assignment (RWA), due to the heavy traffic in network, there will be a need for wavelength assignment to the track in that network. The RWA problem is resolved using the approach of Genetic Algorithm and Ant Colony Optimization Algorithm, wherein up until now is noticed to be better than other optimization algorithms. By using Deep Q-Networks, a type of reinforcement learning, in optimizing the problem in selection of the routing path and wavelength assignment in an optical mesh network.

WDM Long-Haul Networks

Ever-increasing traffic in the long-haul networks called for WDM-based transmission, all the more when compared to the metro and access segments. Further, although the interconnected WDM rings served the needs of long-haul networks for some time, mesh-connected WDM networks appeared as a natural and more effective solution, with enhanced connectivity and resilience against network failures. In this chapter, we first examine the basic design challenges and the candidate node configurations for WRON-based mesh networks. Next we present various offline design methodologies for this class of networks using LP-based and heuristic schemes. Impact of wavelength conversion (WC) in mesh-configured WRONs is also examined by using analytical models, and the gain in bandwidth utilization is assessed in WRONs with full WC as compared to those without any WC. Finally, we examine some online routing and wavelength assignment schemes that can be used for operational WRONs. (143 words)

Dynamic routing and wavelength assignment for efficient traffic grooming

The routing and wavelength assignment (RWA) schemes play an important role in all computer networks. The performance of a computer network and resource utilization largely depend on the RWA. The demand for higher bandwidth is increasing with each passing day, so more efficient RWA schemes need to be devolved to cater the increasing requirements. RWA becomes more challenging for dynamic traffic as the nature and flow of data are not known in advance. In this paper, a dynamic RWA scheme has been proposed for establishing a path in optical networks. The proposed scheme is based upon dynamic conversion sensing algorithm. It can be applied on a number of different network topologies. The proposed scheme is dynamic in nature, which is significantly useful for dynamic traffic grooming. The proposed scheme has been applied on 14 nodes National Science Foundation Network (NFSNet). Simulation results have shown that the blocking probability of this scheme is very low as compared to the existing schemes. So, the proposed dynamic RWA scheme enhances the network efficiency. It is useful for congestion hit networks. The reduced blocking probability in wavelength division multiplexing optical networks leads to better resource utilization and enhanced performance.