IP Fault Localization Via Risk Modeling
Ramana Rao Kompella, Jennifer M. Yates, Albert G. Greenberg, Alex C. Snoeren · 2005
Automated, rapid and effective fault management is a central goal of large operational IP networks. Yet, today’s networks suffer from a wide and volatile set of failure modes, where the underlying fault proves difficult to detect and localize, prolonging the durations of periods of degraded network operation before repair can commence. We introduce a fault localization methodology, based on the use of risk models, and an associated troubleshooting system, SCORE (Spatial Correlation Engine), which automatically identifies likely root causes. We apply the methodology and SCORE to the critical problem that IP operators face today in localizing link failures across IP and optical networks. In experiments conducted on a tier-1 ISP backbone, we find SCORE remarkably effective at this task though it leverages only event data at the IP layers. SCORE thereby bypasses the thorny problem of joining IP and optical event data which are typically described by disparate data models and housed in distinct network management systems. Using risk models constructed from IP and optical topology data, SCORE accurately localizes faults, even those occurring at the optical layer, where event data is unavailable. Moreover, SCORE has proven to automatically uncover inconsistencies in the databases that maintain critical associations between IP and optical networks. As these associations are fundamental to the overall reliability of the network, this is another important benefit. 1