Source-linked AI summary
Virtualized 5G Tesbed using OpenAirInterface: Tutorial and Benchmarking Tests
M. Dória, V. Sousa, A. Campos, N. Oliveira, P. Eduardo, P. Filho, C. Lima, J. Guilherme, D. Luna, I. Rego, M. Fernandes, A. Neto
TL;DR
5G deployment and testing require flexible, cost-effective, and reproducible approaches. The paper provides tutorial material, blueprints, configuration guidance, and throughput benchmarking for a virtualized OAI 5G network using USRP platforms. It reports stable prototyping potential and consistent throughput, while identifying multicell measurements as future work.
Problem
Flexible, cost-effective, and reproducible 5G deployment and testing remain needed, while prior OAI work lacks tutorial material for reproducibility.
Method
The paper presents OAI 5G Core Network and RAN tutorial material, USRP deployment blueprints, configuration discussions, and throughput benchmarking across use cases.
Results
The study finds that OAI and USRP frontends support stable 5G prototyping with consistent throughput and different transmission capabilities, including commercial-phone use.
Takeaways & Limitations
OAI can serve as a practical low-cost platform for 5G prototyping and exploration by industry, startups, small and medium companies, and academic institutions.
Takeaways & Limitations
The evaluation does not yet include multicell measurements, cochannel interference, latency, or error-rate analysis.
Abstract
from arXiv · showhide
The development of 5G and its evolutionary path to 6G brings virtualization as close as possible to the antennas. Native 3GPP systems are now software running at servers boosted by accelerator cards to cope with computationally intense signal processing. Meanwhile, the Radio Frequency (RF) front-end is still proprietary hardware, sheltering specific PHY-layer procedures like passband amplification/modulation. This approach takes advantage of the software's flexibility while keeping the complex microsecond signal processing performance from modern telecommunication systems. This paper provides tutorial material on the Core Network and Radio Access Network of OpenAirInterface (OAI) 5G stack on top of Universal Software Radio Peripheral (USRP) platforms. A set of blueprints showcases OAI's ability to provide a virtualized 5G network with different transmission capabilities and the possibility to use it with commercial mobile phones. Configuration discussions and throughput benchmark analyses follow installation and deployment instructions. Our results show that 5G prototyping using OAI and USRP frontends can lead to good reproducibility and consistent throughput.
1 Introduction
The paper addresses the need for flexible, cost-effective, and reproducible 5G deployment and testing by combining tutorial material with OAI/USRP benchmarking. It positions the work as a practical complement to prior OAI studies that often emphasize theory or results without step-by-step reproducibility guidance.
- The study motivates virtualization and open software by linking reliable operation under varied scenarios and loads to user satisfaction and economic survival.
- The paper provides OAI 5G tutorial material, configuration discussions, throughput benchmarking, and open-source deployment resources for practical prototyping.
- Its blueprints demonstrate virtualized 5G networks with different transmission capabilities and operation with commercial mobile phones.
- The contributions include reproducible OAI/USRP prototyping, throughput analysis, RAN parameter discussions, and public source code for configurations and use cases.
- 1.1 Related Works: Prior OAI research covers virtualization, prototyping, performance testing, resource consumption, scheduling, and applications, but commonly lacks reproducibility-oriented tutorial material.
2 OpenAirInterface (OAI)
OpenAirInterface is an open-source, 3GPP-oriented platform spanning RAN, core-network, and orchestration functions across multiple deployment modes. The paper uses a monolithic standalone RAN configuration because it was more stable for the selected OAI code version.
- 2.1 OAI 5G RAN: The OAI 5G RAN supports 3GPP-compliant standalone and non-standalone architectures, with standalone operation available in split or monolithic modes.
- 2.1 OAI 5G RAN: In split mode, the gNB-CU handles RRC and PDCP, the gNB-DU handles PHY, MAC, and RLC, and the RU handles the radio portion.
- 2.1 OAI 5G RAN: The paper adopts standalone monolithic mode because it offered better stability in the specific OAI code version used.
- 2.3 OAI 5G Core Network: OAI core-network deployments include Minimalist, Basic, and Slicing 5GC modes, while UPF options include SPGW-U-tiny, VPP-UPF, and a production-grade UPF under development.
- OAI supports bare-metal or virtual-machine installation, Docker Compose automation, and cloud-native Helm-based deployment.
3 Installation/Deployment Tutorial
The tutorial automates deployment of the OAI 5GC and gNB through scripts, Docker-based components, and selectable USRP configurations. It also describes host requirements, connection checks, logging, and deployment options for different PRB settings and USRP models.
- The installation requires at least 4 CPU cores, 16 GB RAM, 1.5 GB free Docker storage, Ubuntu 20.04 LTS or newer, Python 3, Docker, and Docker Compose.
- Deployment begins with the 5GC installation script, followed by services.sh to install and deploy the gNB using configurable parameters.
- The core script installs dependencies and certificates, configures packet forwarding and iptables, and performs Docker-related setup operations.
- The gNB script offers n106, n162, and n273 configurations for a USRP N310, plus b106 and b162 configurations for a USRP B210.
- The tutorial verifies connectivity by checking the USRP host connection and following oai-amf logs for 5GC, gNB, and UE status.
- A selected service command builds a Docker gNB image with the specified bandwidth and USRP, illustrating configuration-driven deployment.
4 Test Scenarios
The tests evaluate OAI 5G throughput across carrier bandwidths using a USRP N310, a Motorola G50, and configured gNB parameters. The section details the hardware, SISO settings, frequency calculations, and deployment configurations used to support stable, reproducible experiments.
- 4 Test Scenarios: Throughput tests vary carrier bandwidth from 3.7 GHz using a USRP N310 RF front-end, 7.2x split, and Motorola G50 UE.The physical infrastructure includes a Lenovo SR630 server and two Ettus LP0965 antennas connected through 10Gbps SFP+ interfaces.
- 4.2 gNB Parameters configuration: SISO operation uses one transmit and receive antenna, disables CSI-RS and SRS, and sets min_rxtxtime to 5 ms.The configuration can also disable 256-QAM through force_256qam_off.
- 4.2 gNB Parameters configuration: NR-ARFCN conversion maps frequency values in Hz to channel numbers for configuring dl_absoluteFrequencySSB and dl_absoluteFrequencyPointA.The reference-frequency equation uses FR, ΔFG, FROffs, NR, and NROffs; the paper states that Equations 1 and 2 perform the two conversion directions.
- 4.2 gNB Parameters configuration: The experiments configure 106 PRBs as 40 MHz in n78, with dl_absoluteFrequencyPointA at 3.7 GHz and absoluteFrequencySSB at 3.72 GHz.Downlink and uplink carrier-bandwidth parameters must match.
- 4.2 gNB Parameters configuration: The tutorial also provides open-source step-by-step instructions for configuring Sysmocom SIM cards and enabling UE usage.SIM parameters must be compatible with the configured 5G network, including IMSI, OPc, and Ki.
5 Results
The OAI testbed evaluates spectrum occupancy, laboratory coverage, throughput, RTT, UE-to-UE communication, and distance effects across multiple PRB and bandwidth settings. Results show stable throughput at narrower bandwidths, increasing latency variability with more PRBs, and reduced mean throughput with distance or UE-to-UE forwarding.
- 5.3.1 Channel-Ideal Scenario: 238.60 Mbps average throughput was achieved at 273 PRBs, compared with 207.66 Mbps at 162 PRBs and 135.69 Mbps at 106 PRBs.The 273-PRB configuration reached 297.80 Mbps maximum but had the lowest stability, with 29.02 Mbps standard deviation.
- 5.3.1 Channel-Ideal Scenario: Throughput variation was lower for narrower bandwidths, with more upright CDF curves for 106 and 162 PRBs.The CDF used 1,800 samples from 10 experiments for each PRB configuration.
- 5.3.1 Channel-Ideal Scenario: RTT increased with PRB count, while RTT standard deviation rose from 3.71 ms at 106 PRBs to 10.42 ms at 273 PRBs.The authors attribute the trend to processing requirements that scale with bandwidth allocation.
- 5.3.1 Channel-Ideal Scenario: Packet loss was zero across all experiments, indicating that reported throughput and RTT values were unaffected by network congestion.
- 5.3.2 Two UEs Evaluation Scenario: UE-to-UE communication reached 16.45 Mbps maximum and 14.75 Mbps mean throughput, versus 245 Mbps maximum and 135.69 Mbps mean for gNB-to-UE.The authors suggest uplink limitations and dependence on the uplink/downlink TDD slot ratio constrain the UE-to-UE case.
- 5.3.3 Distance Evaluation: At 12 m, mean throughput decreased to 102.35 Mbps, while maximum bandwidth remained between 119 Mbps and 153 Mbps and variability increased with distance.The results indicate preserved high maximum rates but declining average performance and greater variation at longer distances.
6 Conclusions and Future Works
The paper presents a reproducible, low-cost virtualized 5G testbed using OpenAirInterface and evaluates it through spectrum, throughput, and ping measurements. It concludes that OAI supports stable 5G prototyping while identifying multicell measurements, interference, latency, and error rate as future extensions.
- The study provides network- and user-perspective use cases covering spectrum, throughput, and ping evaluations.
- OAI demonstrated stability and potential for prototyping 5G enhancements on a real low-cost platform.
- Multicell measurements are identified as a next step, including cochannel interference, latency, and error-rate metrics.
Funding
The study reports partial funding, author contributions, data availability, and no competing interests.
- The research was partially funded by Lenovo and CAPES under Finance Code 001.
- Authors contributed across conception, validation, methodology, experiments, writing, supervision, and final approval.
- The authors declare that they have no competing interests.
- Datasets and tutorial materials for reproducing the results are available in an open-source repository.