5GC-Tracer: A Novel Non-Intrusive Distributed Tracing for Enhanced 5G Core Network Observability

The advent of 5G has introduced complex network architectures with stringent Quality of Service (QoS) needs, requiring robust observability solutions for telco operators. This paper proposes 5GC-Tracer, a novel non-intrusive distributed tracing architecture specifically designed for cloud-native 5G Core Network (5GC). Leveraging eBPF technology, our proposal enables application-centric implicit context propagation without requiring any code instrumentation, thus minimizing operational overhead. We conducted real-world testing during the Paris 2024 Paralympic games, demonstrating the effectiveness of 5GC-Tracer in generating and collecting traces from a 5GC system. The experimental results highlight a high success rate in trace collection with low overhead (latency, CPU, memory), and our statistical approach significantly accelerates the localization of anomalies related to QoS degradation across various 5GC network functions.