PRAVEGA: Scaling Private 5G RAN via eBPF/XDP

Abstract

We exploit eBPF+XDP to scale and accelerate software packet processing in (O-RAN compliant) disaggregated 5G RAN (Radio Access Network). We argue that the Central Unit User Plane (CU-UP) component is likely the bottleneck in the 5G RAN user plane data path and therefore focuses on optimizing its performance. We propose an eBPF/XDP-based framework, PRAVEGA, and discuss additional options for further improvements.