Control Plane Performance Benchmarking and Feature Analysis of Popular Open-Source 5G Core Networks: OpenAirInterface, Open5GS, and free5GC

Abstract

This study examines the maturity and state of open-source 5G Core (5GC) networks, with a focus on their support of 5G procedures and Network Function (NF) operations. The research emphasises the importance of optimising the virtualised resource usage of 5GC control plane functions. Given the large set of NF operations and procedures defined for 5GC networks, the study presents a framework that correlates NF operations with 5G procedures, thus facilitating the assessment of 5GC implementations. Furthermore, the study introduces macro-benchmarking and micro-benchmarking approaches to assess the performance of 5GC control plane functions. Our 5G traffic generator is used to generate User Equipment (UE) registration traffic for benchmarking. Macro-benchmarking measures the time taken for a specific number of UEs to complete UE procedures. In contrast, micro-benchmarking analyses the system calls made by the 5GC, presenting their latency and frequency and the processes responsible for generating them. Our findings revealed a direct correlation between macro-level performance and micro-benchmarking results. This suggests that optimising micro-level implementations is crucial for enhancing overall system performance. Moreover, examining micro-benchmarking data can provide valuable insights into how the system’s performance will scale under varying workloads. We use the correlation of benchmarking results to identify performance improvement points in each 5GC and to provide recommendations for software architectural changes that can optimise the usage of virtualised resources. The study includes artefacts and source code to enable replication of the work and results.

Keywords

• 5G
• 5G core networks
• BPF compiler collection tools (BCC Tools)
• free5GC
• Linux Kernel
• load testing