A machine learning approach to mitigating fragmentation and crosstalk in space division multiplexing elastic optical networks.

• We use the Elman neural network (ENN) to obtain an accurate forecast of future based on historical data. • The proposed crosstalk-aware intra-core resource allocation can reduce the inter-core crosstalk. • The proposed horizon-based inter-core spectrum allocation can reduce the fragmentation and improve spectrum utilization. As network traffic is expected to continue to grow at high rates for the foreseeable future, it becomes imperative to introduce space division multiplexing elastic optical networks (SDM-EONs) into the optical transport network. However, spectrum fragmentation and crosstalk present significant challenges that may negatively impact the performance of SDM-EONs. In this paper, we leverage machine learning techniques to enhance the transmission performance of SDM-EONs, and make two contributions. Specifically, we use an Elman neural network to forecast traffic demands, and use a two-dimensional rectangular packing model to allocate spectrum so as to decrease unnecessary spectrum fragmentation (and, in turn, increase resource utilization). We also present a novel spectrum partition scheme to reduce crosstalk. Our evaluation study confirms that the proposed strategy is effective in improving spectrum utilization while reducing blocking probability and crosstalk. [ABSTRACT FROM AUTHOR]