Efficient dynamic routing in Spectrally-Spatially Flexible Optical Networks based on traffic categorization and supervised learning methods

In this paper, we focus on the efficient dynamic routing in Spectrally-Spatially Flexible Optical Networks (ss-fon) realized using Single-Mode Fiber Bundles (smfbs). We study two scenarios – unprotected network (np) and network protected by dedicated path protection (dpp) against a single link failure. For these configurations, we propose a dedicated optimization approach (Enhanced Adaptive Spectral-Spatial Allocation – e-assa), which makes use of the traffic categorization and application of different allocation strategies for different traffic categories. To select beneficial categorization rules, we employ supervised learning paradigm. We show that the selection of a beneficial regression algorithm to support network optimization cannot be performed based only on standard metrics like r2 but some additional measures/experiments are necessary. Then, we carry out extensive numerical experiments in order to tune the approach and to evaluate its efficiency based on the comparison with reference methods. The results prove high efficiency of the proposed optimization framework, which provides low blocking probability and significantly shorter processing time compared to the best of reference methods.