Designing an improved ULCL UPF architecture in the 5G core network

With the rapid growth in demand for mobile devices, the substantial volume of real-time data they generate poses greater challenges for network transmission performance. To support real-time and distributed processing, the 5G core network employs Multi-access Edge Computing (MEC) and Uplink Classifier (ULCL) to route uplink data to nearby edge nodes, reducing end-to-end latency. However, in most real-world deployments, due to Network Function Virtualization (NFV) and cloud-native architectures, network components are often built on virtualized or containerized platforms. As a result, packets must pass through virtual network interfaces, introducing additional delays and processing overhead. Therefore, this study aims to address the limited packet processing performance in virtualized environments and the lack of ULCL support in the open-source project Open5GS. It proposes a ULCL module based on extended Berkeley Packet Filter (eBPF) and Express Data Path (XDP) to enable low-latency, high-efficiency packet steering and close the functional and performance gaps in the current system.