Leveraging Field Data For The Joint Optimization of Capacity and Availability in Low-Margin Optical Networks

With monitoring data collected from a deployed network, we explore ways to improve capacity through rate adaptation with minimal impact on the quality of service. We first perform a large scale statistical analysis of how performance varies in time at the connection level. Then, we quantify the trade-off between capacity and availability stemming from the observed performance variations, and identify a margin sweet-spot where capacity can be close to optimal with minimal cost in terms of availability. Aiming practical implementation, we propose and dry run two rate adaptation mechanisms to jointly optimize capacity and availability based on live monitoring data. With current transponder properties and network infrastructure, we show that a 106% capacity gain is possible while maintaining availability over 99.99% for 95% of connections. Beyond promising numerical results, our methodology can be applied to any network to harmlessly quantify how margin reductions would impact the quality of service, and selecting the most suitable margin-setting approach.