5GC-Observer: a Non-intrusive Observability Framework for Cloud Native 5G System

Telco stakeholders are developing a deeper understanding of cloud native technologies and adopting them faster than few years ago. It is undeniable that migrating legacy telco applications to microservice-based architectures accelerates and facilitates the development of new network services while offering a high level of granularity. However, cloud native raises new operational challenges. In order to achieve an efficient management of network services, new solutions are required to monitor and track widely distributed cloud native network functions while considering their specificity. In this paper, we propose an innovative framework, 5GC-Observer, for the observability of cloud native 5G network services. To the best of our knowledge, no such a solution has been found to date. To achieve its goal, 5GC-Observer relies on the eBPF technology to monitor the network traffic circulating between the 5G core components and report telemetry data. Besides, we leverage a statistical method to detect Quality of Service degradation based on reported telemetry data. Such an approach highlights the richness of the data acquired by our solution and its capability to detect unexpected network-related anomalies. The latter are not detectable through standard observability solutions. Performance evaluation shows that our solution generates low overhead while giving insight into the 5G core system and its internal and external exchanges.