An efficient segment-based defragmentation algorithm for dynamic lightpath reconfiguration in WDM networks

Uncertainty in connection requests arrivals and departures brings forth inefficient resource allocation of online RWA heuristics for dynamic lightpath provisioning. Thereby, periodic re-optimization can play invaluable role in improving resource utilization of WDM networks. This paper investigates network re-optimization carried in two consequent phases: a) finding new wavelength assignment for a subset of lightpaths to minimize spatial fragmentation with the least relocation cost, i.e. lightpath reconfiguration, b) migrating the set of relocatable wavelength paths to the new configuration without incurring undesirable traffic disruption. We assumed sparse wavelength conversion capability where a small set of network nodes are equipped with full-range wavelength convertors to accommodate segment-based defragmentation. The proposed reconfiguration scheme can effectively squeeze optimization cycle-time since only a small set of colliding segments are processed in each cycle. For the migration phase, a Make-Before-Break wavelength relocation methodology is introduced to avoid connection disruption while maintaining minimum relocation cost. The effectiveness of defragmentation process is periodically assessed through the proposed fragmentation ratio metric. Numerical simulation demonstrated significant reduction in blocking probability over a wide range of offered traffic loads.