A flexible offline probabilistic (FOP) algorithm is designed to aggressively accommodate random bandwidth traffic demands in long-haul networks. Compared to algorithms that configure demands according to their maximum bandwidth, the FOP algorithm can save 15% of the spectrum used, accommodating over 99% of the throughput demand.

A probabilistic spectrum Gaussian noise (PSGN) model is proposed to predict the nonlinear noise for random bandwidth traffic in long-haul elastic optical networks. The model reduces the noise estimate 9.1% on average compared to the standard Gaussian noise model applied to the maximum bandwidth.