STRCA: A Lightweight and Accurate Root Cause Analysis System Based on 5G Signalling Trace

Jiang Fang, Jiadong Fu, Jiyan Sun, Liru Geng, Yinlong Liu, Wei Ma

Abstract: The 5G core network (5GC) comprises a multitude of network elements through network function virtualization technology (NFV). A network element fault could seriously degrade the Quality of Service (QoS) for massive users or even interrupt all services, thus resulting in significant economic losses. Furthermore, fault propagation between network elements increases the complexity of localizing the root-cause network element. Therefore, how to accurately localize the root-cause network element in complex 5GC is very difficult. We observe that signalling traces are the key data controlling the communication and collaboration between network elements, which can accurately describe the interaction between network elements. Based on the observation, we propose STRCA, which is the first root cause analysis system based on signalling traces. STRCA first designs a signalling parse module to reconstruct traces from massive signalling packets. STRCA then designs a lightweight and effective anomaly detection module to detect trace anomalies. Finally, based on the detection results, STRCA mines suspicious network element sets and ranks these sets by an innovatively designed metric. Based on a real-world Huawei 5GC, the experimental results demonstrate that the STRCA can accurately and efficiently localize the root-cause network element. STRCA achieves a localization accuracy of 76.3%. Besides, STRCA is able to reconstruct and process 172,060 signalling traces in just 107 s. STRCA can enable accurate localization of 5GC faults with very low time complexity.