An Analytical Approach to Computing Blocking Probabilities in Flexi-Grid Optical Networks

Due to the analytical complexity of routing and spectrum allocation policies – random-fit and first-fit, in flexigrid optical networks, it is necessary to analyze the approximate blocking probabilities. To this end, we present a novel model for reduced link state occupancy, whereby each state of a Markov chain represents the total occupied spectrum slices in a flexi-grid optical link. Using the independence link model, we further develop a method to calculate state-dependent per-class connection setup rates in multi-class flexi-grid optical networks. To alleviate the spectrum fragmentation problem, we extend the analysis with a defragmentation process as a reactive connection reconfiguration scheme, taking into account the probability of observing a link being in a fragmentation state, and show that a reactive defragmentation process can indeed reduce blocking considerably. We show that the defragmentation process is most effective under the medium load condition as analysed in a large scale flexi-grid optical networks. Through various numerical and simulation results we show that our model, with or without defragmentation, can be efficiently used to calculate accurate blocking probabilities in multi-class flexi-grid optical networks, as the approximate results match closely with the exact and simulation results under different scenarios, including varying link capacities, classes of demands, and traffic loads.