FALCON: Fault Management via Alarm Warehousing and Mining

The ability to manage faults in large scale networks is of vast importance for successful and eeective network management operations. In this paper, we describe FALCON, a project underway at AT&T Labs-Research focusing on fault management in IP networks. FALCON's goal is to automate various fault management tasks through warehousing and mining technologies. We describe FALCON's architecture and comment on its components. We present ALVIS, an alarm navigation and visualization tool developed to explore alarms generated by AT&T's IP backbone. ALVIS displays alarms of multiple network elements as a set of state time series in a single picture, thereby revealing interesting structure and anomalies in the data. We also describe KALHAS, FALCON's data mining agent, currently under development, and outline our vision and current research agenda towards fault management automation through successful development and deployment of state of the art incremental mining and stream computation techniques.