AHSWN Home • Issue Contents • Forthcoming Papers

Causality Graph Construction of Fault Alarms for Wireless Sensor Networks
Yuhan Jing, Lei Zhang, Cong Liu, Jing Wang, Wei Li, Bo He, Qi Qi and Jingyu Wang

The running time statuses of IoT (Internet of Things) systems are monitored from their wireless sensor networks. When faults occur in the IoT system, their influence may spread to the whole system and trigger a large number of alarms on sensors, forming alarm storms. In order to locate the root cause of faults and reduce the number of alarms to be handle, we proposed an approach to constructing the causality graph for alarms to extract the key information of alarm storms in this paper. The long-term propagation mode of alarms during faults is analyzed to build a corresponding directed acyclic graph which expresses the causal relationships between different types of alarms in the system. Two datasets of alarms collected from China Mobile are used in experiments of extracting summaries for alarm storm cases. With ensuring a high recall of 0.96, the compression rate has reached 1.504.

Keywords: alarm compression; system faults; alarm storm summary; causality graph; artificial intelligence for IT operations; internet of things; wireless sensor networks

Full Text (IP)