On the comparative efficiency of non-disruptive defragmentation techniques in flexible-grid optical networks

The presence of dynamic traffic in optical networks can lead to spectrum fragmentation, which significantly reduce the network performance. This problem is exacerbated in flexible grid optical networks, since the heterogeneity of channel spectral widths increases the misalignment of the available slots along the optical connections. To overcome this problem, several techniques have been devised for providing defragmentation in these networks. Spectral defragmentation aims to create large contiguous free areas in optical spectrum through rerouting or spectral reassignment of existing connections. In this paper, we revisit the push-pull and hop tuning techniques, assessing them over an unified framework under equal conditions to properly evaluate their comparative efficiency in releasing network resources. This framework also includes the re-planning technique as a lower bound benchmark for the push-pull and hop tuning. To this effect, we propose Integer Linear Programming (ILP) models and heuristic algorithms to study the effectiveness of these techniques and present a performance analysis based on spectrum usage. The relative performance of the different defragmentation techniques was validated under varying network scenarios, including an assessment of how different planning methods and policies impact the effectiveness of each defragmentation technique. Aside from benchmarking the potential spectrum efficiency gain in each non-disruptive defragmentation technique, this analysis also revealed that the gains associated with them are greatly influenced by how the networks were originally planned.