Fast traffic processing in multi-tenant 5G environments: A comparative performance evaluation of P4 and eBPF technologies

• Published in: Engineering science and technology, an international journal Vol. 52; p. 101678
• Main Authors: Gallego-Madrid, Jorge, Molina-Zarca, Alejandro, Sanchez-Iborra, Ramon, Ortiz, Jordi, Skarmeta, Antonio F.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2024; Elsevier
• Subjects: extended Berkeley Packet Filter (eBPF); Quality of Service (QoS); Quality of Service (QoS); 5G; P4; extended Berkeley Packet Filter (eBPF)
• ISSN: 2215-0986; 2215-0986
• DOI: 10.1016/j.jestch.2024.101678

Although the softwarization of network infrastructures through the use of Software Defined Networking (SDN) and Network Function Virtualization (NFV) has set the foundations of future communication architectures, the efficient handling of high throughput traffic while maintaining latency requirements still remains a challenge. In this work, we explore two arising technologies that aim at reducing networking tasks’ latency while dealing with high levels of traffic volume, namely, Programming Protocol-independent Packet Processors (P4) and the extended Berkeley Packet Filter (eBPF). We present a review of the latest advances in the use of both technologies and we provide a discussion on their advantages and disadvantages. As the main contribution of the paper, we showcase an extensive performance evaluation of these technologies under different traffic conditions. To do so, we implement a fast traffic processing network function operating in a real 5G Stand Alone (SA) network. Obtained results confirm, as expected, the high performance attained using dedicated hardware programmed by P4, in contrast to eBPF-based solution’s poorer results while handling similar throughputs. Nevertheless, eBPF allows similar packet-processing times than P4, therefore qualifying it as a perfectly scalable solution on commodity hardware even as a virtual function, which paves the way for the realization of autonomous, flexible and cost-effective next-generation network infrastructures.