Predictive Analytics for Anomaly Detection and Failure Prediction in Complex Core Routers
Krishnendu Chakrabarty · 2018
Prognostic diagnosis is desirable for commercial core router systems to ensure early failure prediction and fast error recovery. The effectiveness of prognostic diagnosis depends on whether anomalies can be accurately detected before a failure occurs. However, traditional anomaly detection techniques fail to detect “outliers” when the statistical properties of the monitored data change significantly with time. This talk will describe recent advances in using time-series data analysis to detect anomalies through the real-time monitoring of key performance indicators in core routers. The speaker will describe the design of a changepoint-based anomaly detector ad health-status analyzer that first detects changepoints from collected time-series data, and then utilizes these changepoints to detect anomalies. A clustering method is first used to identify a wide range of normal/abnormal patterns from changepoint windows. Symbolic aggregation approximation and moving-average-based trend approximation are utilized to encode complex time series. Hierarchical agglomerative clustering and sequitur rule discovery are then to learn important global and local patterns. A comprehensive set of experimental results will be presented for data collected during 30 days of field operation from over 20 core routers deployed by customers of a major telecom company.