Date of this Version
Mehr, SY (2020), "An ANNs Based Failure Detection Method for ONOS SDON Controller," Poster presentation, UNL Research Fair, Spring 2020, University of Nebraska-Lincoln.
Network reachability is an important factor of an optical telecommunication network. In a wavelength-division-muliplexing (WDM) optical network, any failure can cause a large amount of loss and disruptions in network. Failures can occur in network elements, link, and component inside a node or etc. Since major network disruptions can caused network performance degradations, it is necessary that operators have solutions to prevent such those failures. This work examines a prediction model in optical networks and propose a protection plan using a Machine Learning (ML) algorithm called Artificial Neural Networks (ANN) using Mininet emulator. ANN is one of the best method which applied for failure prediction and identification in optical networks. The simulation result show the advantages of using ANN method. Also, it has proved that the prediction accuracy was greater than 90 %. on the ONOS controller.
• An ANN prediction method can predict the board failure in a WDM network. • ANN is proven to be very efficient as a classifier (Figure 8). • Services can be protected from data loss before a network failure occurs. • Since ANN has nonlinear nature, then it is more suitable for failure prediction in Optical Network devices. • Environmental temperature was the first labels which helped for board failure. It was about 78.38%. • According to the experiment data, the accuracy of the ANN prediction method was 95.59% on average, which meant the failure state of 96% of the boards could be correctly predicted. • Another advantage of ANN implementation scenario is monitoring optical performance. This features is not accessible in SVM. • The results of performance monitoring might be used in designing optical devices. • Estimation maximization of the optical link capacity can be achieved in ANN method. Such as a linear or nonlinear optical signal to noise ratio (OSNR).