SCORE
Spatial Correlation for Fault Localization
Spatial Correlation for Fault Localization
Project Overview
Operational IP networks are intrinsically exposed to a wide variety of faults and impairments. These networks are large, geographically distributed and constantly evolving, with complex hardware and software artifacts. After much effort, network hardware has been designed and implemented, the protocols controlling the network have been designed (often in compliance with published standards), and the associated software implemented. The rapid pace at which each of these tasks are performed often produces operational artifacts far more richer than that can be approximated in a lab. The goal of this project is to devise mechanisms that assist the network operator in rapid fault localization in order to cut down the mean-time-to-repair for the particular failure scenario.The main technique for fault localization is spatial correlation of temporally correlated events. For example, if three IP links fail simultaneously, there is a good chance that all the three IP links fail because of the same root cause. These link down events would be temporally correlated, if they indeed have the same root cause. Once we obtain these observations through any of the existing monitoring mechanisms such as SNMP traps or syslogs etc., we can spatially correlate the optical paths associated with these links. Any shared risk that exists between these links is quickly exposed with the help of simple inference algorithms.
People
Faculty/Collaborators :- Jennifer Yates (ATT Labs -- Research)
- Albert Greenberg (ATT Labs --
Research),
- Alex C. Snoeren (UCSD)
Papers
Detection and Localization of Network Blackholes,Ramana Rao Kompella, Alex Snoeren, Jennifer Yates, Albert Greenberg,
to appear in Proceedings of IEEE Infocom, ,
(Infocom) Anchorage, AL, May 2007 [pdf] [ps]
Cross-Layer Visibility as a Service,
Ramana Rao Kompella, Albert Greenberg, Jennifer Rexford, Alex C. Snoeren, Jennifer Yates,
to appear in the Proceedings of Fourth ACM Workshop on Hot Topics in Networks,
(HotNets), College Park, MD, November 2005 [pdf] [ps] Conference Talk Slides [pdf]
IP Fault Localization via Risk Modeling,
Ramana Rao Kompella, Jennifer Yates, Albert Greenberg, Alex C. Snoeren,
to appear in the Proceedings of Second ACM/USENIX Symposium on Networked Systems Design and Implementation,
(NSDI), Boston, MA, May 2005 [pdf] [ps] Conference Talk Slides [pdf]