Perseverance-Aware Traffic Engineering in Rate-Adaptive Networks with Reconfiguration Delay

Abstract

Expensive optical fibers provide connectivity for wide-area networks. Nowadays, the fibers are operated in a much conservative manner. By adaptively reconfiguring the fibers to exploit its signal quality, a recent proposal demonstrates a significant increase of optical link capacity. Such a reconfiguration currently accompanies a non-ignorable delay, during which the reconfigured link is not accessible, and the mentioned approach trades off the final throughput with the induced churn during the transition. This scheme can result in high traffic disturbance during the reconfiguration.To overcome the drawback of the simple churn-based update, we study the rate adaptation planning (RAP) problem under reconfiguration delay. We propose a multiple step planning with perseverance constraints. This approach leads to a smoother transition, but the optimal plan is shown NP-hard without constant factor approximation (unless P= NP). Therefore, we develop an efficient LP-based heuristic algorithm. Extensive simulations show that the algorithm gives 40-50% higher throughput than the no-adaptive-link case. Also, the transition is much smoother: the resulting traffic fluctuation is only 40-50% of the existing churn-based approach.