Artificial Neural Network-Aided Multiclass Service Provisioning and Prioritization in EONs

With the emergence of various applications with diverse transmission requirements, optical networks must support multiclass service provisioning and prioritization. This paper studies serving multiclass requests in elastic optical networks by defining appropriate controlling parameters (CPs), which can be learned by an artificial neural network (ANN). These CPs can be tuned to allocate resources according to the service class priorities. We propose two routing, modulation, and spectrum assignment (RMSA) mechanisms working together to establish data and flow-oriented connections. Two physical layer impairment methods are leveraged to realize connection provisioning by considering the quality of transmission limitations. The RMSA methods are equipped with four CPs to handle the time and frequency resource consumption competition between request classes. An ANN-based regression model is used to determine the value of each CP for preset request classes’ blocking ratios. Exhaustive simulations are performed and analyzed to scrutinize the impact of each CP on the network status. Simulations reveal the effectiveness of ANN decision-making in policy imposition of allocating resources to each request class.