Performance Management of Optical Transport Networks through Time Series Forecasting

As the importance of the Internet increases, developing technologies for data transmission at higher bit rates became as important as challenging. Optical Transport Networks (OTN), based on Wavelength Division Multiplexing (WDM), emerged to meet this need. ITU has standardized OTN, including its management functions. Network elements in OTN store performance monitoring information as time series of performance events, which could be explored to enable more efficient preventive maintenance and better resource provisioning. With the use of time series forecasting models, faults could even be avoided, along with Service Level Agreement breaches. In this work, we propose the use of time series forecasting in performance monitoring of OTN, investigating how three forecasting methods behave in scenarios of stability and degradation in network performance and how the number of monitoring intervals affects their accuracy. In accordance with results we state recommendations for application of tested time series forecasting methods in managing OTN.