Recently, sudden traffic spikes due to cyber attacks, flash crowds, and heavy content delivery traffic
have often caused unpredictable congestion. Moreover, the congestion in large-scale networks can
result in service outages that inflict a lot of damage on customers. To provide quality network service
without such network anomalies, it is required to analyze network data, proactively detect the
anomalies, and find root causes. We propose two technologies for performing such proactive network
operation; 1) network fault diagnosis technology based on network log tensor factorization (LTF), and 2)
predictive-based traffic engineering, i.e., traffic-route control technologies using the prediction of traffic
variations based on the diagnosis. We examine these technologies through actual network data.
Please click the thumbnail image to open the full-size PDF file.