Efficient Discrete Rate Assignment and Power Optimization in Optical Communication Systems Following the Gaussian Noise Model

Computationally efficient heuristics for solving the discrete-rate capacity optimization problem for optical fiber communication systems are investigated. In the Gaussian noise nonlinearity model regime, this class of problem is an NP-hard mixed integer convex problem. The proposed heuristic minimizes the number of calls required to solve the computationally intensive problem of determining the feasibility of proposed discrete rate allocations. In a live system, optimization using this algorithm minimizes the number of potential discrete rate allocations tested for feasibility while additional discrete system capacity is extracted. In exemplary point-to-point links at 50 Gbaud with 50 Gb/s rate steps, the mean lost capacity per modem is reduced from 24.5 Gb/s with truncation to 7.95 Gb/s with discrete-rate optimization. With 25 Gb/s rate steps, the mean lost capacity is reduced from 12.3 Gb/s to 2.07 Gb/s. An unbiased metric is proposed to extend the capacity optimization objective from point-to-point links to mesh networks. A gain of 13% in distance-times-capacity metric is obtained from discrete-rate optimization with 50 Gb/s rate steps, and a 7.5% gain is obtained with 25 Gb/s rate steps.