Enhancing the solvability of network optimization problems through model augmentations

Intensive research effort has been dedicated to tackle multi-hop network problems. Joint consideration across multiple layers is required to achieve optimal performance. The general trend in solving these problems is to develop strong mathematical programming formulations that are capable of providing near-optimal solutions to practical-sized problems. For the class of problems studied, we show that a traditionally formulated model turns out to be insufficient from a problem-solving perspective. When the size of the problem increases, even state-of-the-art optimizers cannot obtain an optimal solution because of running out of memory. In this work, we show that augmenting the model with suitable additional constraints and structure enables the optimizer to derive optimal solutions, or significantly reduce the optimality gap, which were previously elusive given available memory restrictions.