Investigation Of Spectrum Defragmentation Algorithms In Elastic Optical Networks

Nowadays, flexible-grid elastic optical networks (EONs) have attracted intensive research interests for the agile spectrum management in the optical layer. Meanwhile, due to the relatively small spectrum allocation granularity, spectrum fragmentation has been commonly recognized as one of the key factors that can deteriorate the perfor-mance of EONs. Quite some research have been conducted to alleviate the negative effects of spectrum fragmentation on the network efficiency.However, the challenges for solving this problem lie in several aspects. The first is the scalability issue. A lot of existing research have not conducted tractability analysis to the algorithms they design. In reality, the defragmentation problem is in large-scale and needs efficient algorithms to solve in reasonable time. Secondly, besides the net-work efficiency, the cost incurred by defragmentation, such as defragmentation latency and traffic disruption, should also be taken into consideration, since these factors will determine whether a defragmentation strategy can be used in practice. However, as far as we know, there are no previous work to quantify these metrics. The last but not the least, we should not only care about how to consolidate the spectrum but also be aware of the feasibility of the proposed consolidation scheme. The meaning of feasibility refers to the fact that as there might be dependency between traffic migrations, achiev-ing desired consolidation might incur considerable traffic disruptions. Above all, in this study, we focused not only on the defragmentation performance on reducing blocking probability but also on minimizing the defragmentation latency, traffic disruptions.First, we proposed the procedure for sequential spectrum defragmentation, includ-ing defragmentation object selection algorithms, fragmentation-aware routing and spec-trum allocation algorithms, traffic migration algorithms. The simulation results show that our algorithms can achieve very good performance on reducing blocking probabil-ity while minimizing the traffic disruptions.Secondly, we modeled the defragmentation latency and traffic disruption based on software-defined EONs architecture. The quantitate analysis have been given on the de-fragmentation latency and traffic disruption of sequential and parallel defragmentation respectively.Thirdly, we studied the parallel defragmentation algorithms. We proved the re-active parallel defragmentation problem is a strongly NP-hard problem, which needs the optimal solution. So we turned to the proactive defragmentation problem, which is proven to be an APX-hard problem by us. And we proposed an efficient Lagrangian-relaxation based heuristic algorithms to solve it. The simulation results show that our algorithm converges fast.By comparing the simulation results on sequential and parallel defragmentation, we found that sequential defragmentation can make the spectrum more consolidated and thus reduce more blocking probability. By contrast, parallel defragmentation sac-rifices the defragmentation performance while achieves minimum latency and traffic disruption. The service provider can select either sequential or parallel defragmenta-tion strategy according to its requirements.