Regression-based fragmentation metric and fragmentation-aware algorithm in spectrally-spatially flexible optical networks

Abstract Spectrally-spatially flexible optical networks (SS-FONs) are seen as a next frontier in optical backbone networks that allow supplying demanded high-capacity transmission. In SS-FONs, signals are co-propagating in spatial modes of suitably designed optical fibers, e.g., in the bundles of single-core single-mode fibers. Despite significant fiber capacity, SS-FONs operate on a flexible (elastic) grid which allows for assigning an adjustable amount of spectrum resources according to the requested bit-rate. The full potential of SS-FONs’ spectral and spatial flexibility can be exploited when nodes are equipped with switching devices enabling lane changes, i.e., the devices that support arbitrary switching between input and output spatial modes connected to the node. However, before the SS-FON will reach maturity and become ready for commercial applications, several crucial issues need to be solved. In this paper, we study the fragmentation problem for dynamic traffic in SS-FONs with lane changes. We propose a novel weighted fragmentation metric that accounts for vertical and horizontal fragmentation in the considered scenario. The machine learning regression model is created and solved to obtain the best weights combination that minimizes the network fragmentation. We run experiments on the representative network topology using our developed fragmentation-aware algorithm showing that the proposed metric and assigned fiber weights result in network fragmentation decrease. As a consequence, the proposed solution allows for bandwidth blocking probability reduction when compared to the reference methods. Finally, we discuss several optimization strategies that decrease the computational complexity of our algorithm.