Study compares OpenAirInterface and srsRAN across features, throughput and real app workloads

The comparison at a glance

A paper that surfaced on the Hacker News front page tackles a question anyone building a private 5G network eventually faces: do you build on OpenAirInterface or srsRAN? The study, submitted to arXiv on 26 August 2026 by Maria Katarine Santana Barbosa, compares the two prominent open-source implementations of the 5G Radio Access Network (RAN) from two directions: what each platform supports, and how each actually performs under realistic traffic.

The gap in earlier evaluations

According to the paper, existing assessments of OAI and srsRAN tend to be narrow in scope. Some confine themselves to a single testbed, others run end-to-end measurements, and a few attempt only partial comparisons between the two stacks. The authors argue that a meaningful evaluation has to vary several dimensions at once: the software-defined radios (SDRs) serving as the radio hardware, the RAN configuration, the applications running on top of the network, and the number of real user devices, or UEs, connected at any given time.

Because actual deployments differ across exactly those dimensions, results from a fixed, single-purpose setup do not necessarily transfer to another environment. That mismatch between how the platforms are usually evaluated and how they are actually deployed is the problem the paper sets out to address.

Two tracks of analysis

The evaluation splits into a qualitative and a quantitative half.

The qualitative side catalogues which features each platform supports and how flexible each is to deploy, the practical questions that determine whether a stack fits a given environment before any throughput is measured.

The quantitative side then measures three things. First, how each platform handles Radio Resource Control (RRC) setup, the signalling procedure a device goes through when it joins the network. Second, how closely observed data rates track the theoretical maximum for a given configuration. Third, how the platforms cope with real application workloads that place different quality-of-service demands on the network.

A workload mix built for contrast

The application layer is where the study becomes practical. Video on Demand (VoD) serves as the data-rate-oriented workload, stressing sustained throughput rather than reaction time. Live Streaming (LS) and Cloud Gaming (CG) are the latency-sensitive picks, chosen because they need the network to keep up in real time. The authors say the three workloads together provide a comprehensive view of what each platform can deliver.

What the abstract leaves open

The public abstract lays out the methodology but not the findings. It does not state which platform came out ahead on any metric, or by how much; the concrete numbers on setup behaviour, throughput and latency sit in the full paper. What the abstract does establish is the shape of the comparison, and that shape is itself useful for teams planning evaluations of their own.

Why it matters

Private 5G has been a persistent enterprise pitch, and open-source RAN stacks are what make it feasible at small scale. As the paper notes, OAI and srsRAN have become increasingly relevant for industry and academic research, rapid prototyping and low-cost private deployments. Until now, choosing between them has relied heavily on scattered, single-setup studies. A comparison that spans feature support, deployment flexibility, signalling behaviour, theoretical-rate compliance and mixed-QoS real applications gives operators and researchers a common reference point, and a reusable template for benchmarking their own hardware and configurations before committing to a stack.