Intra-Data Center Network Optimization by Means of Application-Based Traffic Modeling and Optical Circuit Switching

In this paper, we investigate a hybrid electro-optical Intra-Data Center Network based on a service-customized Long-Range Dependence traffic model specific to machine-generated data. We show that different application classes, typically hosted by data centers, possess distinct and highly varying behaviors, which need to be considered for an efficient architecture design and optimization. In this regard, using a statistics-based simulative demonstration, we show the benefits brought by optical circuit switching, namely due to its potential in handling bulky and highly intensive data traffic peculiar to some applications. Owing to the fact that cost savings represent another essential target of network optimization, we also quantify the gains related to capital and operational expenditures specific to the new hybrid scheme.