eBPF and XDP Technologies as Enablers for Ultra-Fast and Programmable Next-Gen Network Infrastructures

Current and next-generation services present a challenging set of requirements that network infrastructures need to cope with. Although Software Defined Networking (SDN) and Network Function Virtualization (NFV) techniques have laid the foundations of future communication architectures by providing flexibility and robustness against operational changes, the reduction of the networking tasks’ latency still remains as a challenge. This is specially crucial for enabling efficient resource management in fog/edge/cloud computing scenarios. In this line, the extended Berkeley Packet Filter (eBPF) has appeared as a promising solution to reduce these delays and to provide advanced programmability capabilities to the infrastructure. In this work, we explore eBPF and the eXpress Data Path (XDP) as key enablers for the design of next-generation networks as well as their advantages and limitations. Also, we present a case study in which an eBPF/XDP network function has been developed and integrated within a real 5G network adopting a Mobile Edge Computing (MEC) approach, with the aim of applying diverse Quality of Service (QoS) policies to different multi-tenant 5G traffic flows. The obtained results show how the proposed solution outperforms a baseline approach powered by iptables, achieving a reduction of around 40% in the packet processing time. The proposed solution paves the way for the development of novel virtual network and resource management functions.