Bandwidth defragmentation in dynamic elastic optical networks with minimum traffic disruptions

Mingyang Zhang,Weiran Shi,Long Gong,Wei Lu,Zuqing Zhu

DOI: https://doi.org/10.1109/ICC.2013.6655165

2013-01-01

Abstract:Bandwidth defragmentation, i.e., the operation to reconfigure existing connections for making the spectrum usage less fragmented and less misaligned, has recently been recognized as one of the most important features for elastic optical networks (EONs). In this paper, we propose a novel comprehensive bandwidth defragmentation algorithm that considers the problems of 1) When to defragment? 2) What for defragment? and 3) How to defragment? jointly. The proposed algorithm accomplishes defragmentation through proactive network reconfiguration that only reroutes a portion of existing connections. In each defragmentation operation, we first choose the connections to reroute using a selection strategy, then determine how to reroute them with the defragmentation based routing and spectrum assignment (DF-RSA), and finally perform rerouting with best-effort traffic migration to minimize traffic disruptions. Simulation results indicate that in order to make the bandwidth blocking probability (BBP) comparable with that from the Greenfield scenario (100% rerouting all the time), the proposed algorithm only needs to reroute ~30% existing connections. The simulations also demonstrate that the traffic disruption percentages are less than 1% for defragmentations with 30% rerouting and can be further reduced to within 0.25% by adding a move-to-vacancy (MTV) approach in the traffic migration.