PaperlayerView exact revision on arXiv

Source-linked AI summary

Open-Source 5G RAN Platforms: A Dual Perspective on Performance and Capabilities

Maria Katarine Santana Barbosa, Iasmin Gomes, Vinícius Melo, Kelvin Lopes Dias

arXiv:2608.25820v1cs.NI

TL;DR

Existing evaluations of OAI and srsRAN rarely cover different SDRs, RAN configurations, real applications, and multiple real UEs together. This paper compares the platforms qualitatively and quantitatively using a low-cost standalone prototype, finding broadly similar overall performance with platform- and scenario-specific differences.

  • Problem

    Evaluations of OAI and srsRAN remain limited across different SDRs, RAN configurations, real applications, and numbers of real UEs.

  • Method

    The paper combines qualitative feature analysis with quantitative evaluation of OAI and srsRAN using a low-cost standalone prototype, multiple SDRs, configurations, UEs, and application workloads.

  • Results

    Overall performance was similar: OAI achieved the best RRC Setup results, while srsRAN was closest to theoretical throughput and showed stronger consistency across application scenarios.

  • Takeaways & Limitations

    The comparison identifies complementary platform strengths, including OAI's 120 kHz SCS support in FR2 and srsRAN's documentation and performance consistency.

Abstract

from arXiv · show

Open-source implementations of the fifth generation (5G) Radio Access Network (RAN), such as OpenAirInterface (OAI) and srsRAN, have become increasingly relevant for industry and academic research, rapid prototyping, and low-cost private 5G deployments. The state of the art presents different evaluation scenarios for both platforms; however, these are often limited to individual testbed evaluations, end-to-end studies, or partial comparisons between them. In particular, there is a lack of evaluations considering different Software-Defined Radios (SDRs), RAN configurations, user applications, and numbers of real UEs. To address this gap, this paper presents a comparative evaluation of OAI and srsRAN. The study includes a qualitative assessment of supported features and deployment flexibility, followed by a quantitative performance analysis covering radio resource control setup procedures, compliance with theoretical data rates, and real application workloads with diverse quality of service requirements. Specifically, we consider Video on Demand (VoD) for data rate evaluation, Live Streaming (LS) and Cloud Gaming (CG) as latency-sensitive applications. These workloads provide a comprehensive view of each platform capability.

I. INTRODUCTION

5G RAN research has expanded from proprietary, monolithic systems toward virtualized open-source platforms, but comparative evaluations across radios, configurations, real UEs, and applications remain limited. The paper addresses this gap through qualitative platform analysis and quantitative prototype evaluation.

  • Functional RAN splits and standardized interfaces enabled virtualized deployments on Commercial Off-The-Shelf hardware, supporting open-source platforms such as srsRAN and OAI.
  • Few studies evaluate RRC setup across different USRPs and multiple real UEs.
  • The study compares OAI and srsRAN qualitatively, then evaluates a low-cost standalone 5G prototype with two USRPs, multiple radio configurations, and real applications.
  • The evaluation covers VoD, LS, CG, and iPerf3 while analyzing control-plane setup, theoretical data rates, SDR differences, and application behavior.

II. OPEN SOURCE 5G RAN PLATFORMS

The paper compares open-source 5G RAN platforms by general characteristics, deployment support, and implemented features. OAI offers broad deployment flexibility and experimental FR2 support, while srsRAN emphasizes simpler installation and configuration with selected Release 17 capabilities.

  • The platform comparison covers 3GPP release, installation and configuration difficulty, programming language, virtualization, radio features, functional split, DPDK, duplexing, slicing, and NTN support.
  • OAI supports up to 100 MHz bandwidth, TDD/FDD, 256-QAM, 2x2 MIMO, and deployment through bare metal, Docker, or Kubernetes.
  • OAI experimentally supports 120 kHz SCS in FR2, while detailed parameter configuration remains challenging despite extensive documentation.
  • srsRAN supports FR1 FDD/TDD, bandwidths up to 100 MHz, selected 3GPP Release 17 features, split 7.2, and up to 4x4 MIMO.
  • srsRAN generally provides simpler installation and configuration, with source-build or Ubuntu-package installation options and a comprehensive parameter reference.

III. 5G NETWORK PROTOTYPE

The proposed standalone 5G prototype connects services, the 5G core, the RAN, and mobile users to evaluate OAI and srsRAN under diverse application workloads. The deployment uses Docker-based software components and SDR-based radio hardware.

  • The testbed architecture is organized into service, 5GC, RAN, and UE layers.
  • The service layer provides iPerf3, VoD, LS, and CG workloads for performance evaluation under diverse demands.
  • The 5GC handles authentication, registration, PDU session establishment, and data-network interconnection through the N6 interface.
  • The RAN implements 3GPP functional split Option 8, with OAI and srsRAN deployed in Docker on an x86 server and radio functions assigned to SDRs.
  • The testbed specification includes the virtual infrastructure, software versions, mobile device, and Docker networking between the 5GC and gNodeB.

A. Radio Configuration

The evaluation fixes modulation and subcarrier spacing while varying bandwidth and MIMO configurations. A TDD slot pattern allocates time between downlink and uplink, directly shaping achievable data rates.

  • The achievable single-carrier data rate depends mainly on modulation, PRBs, and the number of layers.
  • The experiments use fixed 256-QAM and 30 kHz SCS across 20 MHz single-layer, 40 MHz single-layer, and 20 MHz 2×2 MIMO configurations.
  • The 2×2 MIMO configuration is not combined with 40 MHz bandwidth because that operating mode requires a 46
  • The TDD pattern is [D D D D D D D F U U], with the flexible slot assigned eight symbols to downlink and four to uplink.
  • Because uplink and downlink share frequency through time-separated slots, the TDD pattern directly affects achievable data rates.

IV. PERFORMANCE EVALUATION

The evaluation compares RRC setup time, theoretical-throughput proximity, and performance under real workloads including VoD, LS, and CG.

  • The evaluation covers control-plane RRC setup time and data-plane performance against theoretical rates and real application workloads.Real workloads include VoD, LS, and CG, each with different throughput or latency requirements.

A. Case 1: RRC Setup procedure

This case measures the time required to complete RRC Setup across platforms, radio configurations, SDRs, and multiple real UEs. C1 produced the best average result, while overall configuration behavior was similar.

  • A. Case 1: RRC Setup procedure: 87.7 ms was the best average RRC Setup time, achieved by C1 with both srsRAN and OAI.The experiment used four UEs and ten connection-disconnection cycles per UE.
  • A. Case 1: RRC Setup procedure: 85.3 ms was the best isolated result, recorded by OAI with configuration C3 on the B210 board.
  • A. Case 1: RRC Setup procedure: 15.2% higher than the best result was the average time for OAI with configuration C2 on the N310.Overall, the configurations exhibited similar behavior for this metric.

B. Case 2: Theoretical Throughput

The theoretical-throughput case measures end-to-end downlink and uplink capacity against 3GPP-derived targets across configurations and SDRs. srsRAN was generally more consistent, while OAI’s results varied substantially, especially in C3 MIMO tests.

  • B. Case 2: Theoretical Throughput: 18–24% higher downlink throughput than OAI was achieved by srsRAN in C1 for B210 and N310, respectively.
  • B. Case 2: Theoretical Throughput: 45.7% and 43.8% of srsRAN throughput were achieved by OAI in C3 with B210 and N310, respectively.The passage attributes OAI’s C3 performance drop mainly to instability during MIMO tests.
  • B. Case 2: Theoretical Throughput: Approximately 70% of expected throughput was reached by srsRAN in downlink scenarios.
  • B. Case 2: Theoretical Throughput: 5–49% of theoretical capacity was reached in uplink across all scenarios.srsRAN remained more consistent except for B210 with C3, which was unstable and could not be completed.
  • B. Case 2: Theoretical Throughput: srsRAN maintained efficiency across different SDR boards, while RAN configuration parameters had the main impact on spectral efficiency.

C. Case 3: Performance under Real Application Workloads

Real-application evaluation covers workloads with different throughput and latency requirements, using VoD traffic generated by simultaneous UEs. srsRAN showed more stable VoD quality under load, while OAI was more configuration-sensitive.

  • 1) Video on Demand (VoD):: Four simultaneous UEs requested VoD chunks, with each UE downloading about 60 chunks lasting 10 seconds.The Android application requested Big Buck Bunny 60fps/1080p and reported network metrics to the server.
  • 1) Video on Demand (VoD):: 1.5 seconds was srsRAN’s median VoD loading time, compared with more than 2.8 seconds for OAI for 50% of video chunks in C1.The comparison was particularly pronounced with the N310.
  • 1) Video on Demand (VoD):: C2 provided the most balanced VoD performance across both stacks and SDRs, with higher throughput and lower loading times than other configurations.
  • 1) Video on Demand (VoD):: srsRAN maintained better QoS stability under four-UE load, whereas OAI performance was more sensitive to configuration parameters.

2) Live Streaming:

The Live Streaming evaluation measures latency across OAI and srsRAN platform combinations using different SDRs and configurations. Overall, srsRAN generally provides lower and more stable latency, while OAI varies substantially by configuration.

  • Live Streaming: Live Streaming evaluates network latency for each platform combination.The evaluation uses an open-source broadcasting and playback workflow with measurements collected at the UEs.
  • Live Streaming: 33.91% higher latency and connection loss with the B210 occurred for OAI in C1 compared with srsRAN.For srsRAN, the B210 achieved about 120 ms mean latency, while N310 was only 2.5% worse.
  • Live Streaming: 108 ms for OAI and 109 ms for srsRAN were the mean latencies with N310 in C2, making it the most suitable SDR for that configuration.OAI with B210 was the worst C2 case, reaching 320 ms maximum and 120 ms mean latency.
  • Live Streaming: OAI with N310 in C3 reached 485.09 ms mean latency and unstable behavior, whereas srsRAN recorded 113 ms on B210 and 119 ms on N310.The figure compares LS latency across configurations and 5G stacks.

3) Cloud Gaming:

The Cloud Gaming workload evaluates 5G latency while three UEs simultaneously play an offline game. srsRAN maintains similar latency across configurations, whereas OAI is generally unstable except with C2 and B210.

  • Cloud Gaming: Cloud Gaming streams Brawlhalla at 60 FPS, 20 Mbps, and 1080p using Sunshine, Steam Remote Play, and Moonlight.The local server and client setup isolates latency generated by the 5G environment.
  • Cloud Gaming: Three UEs played Brawlhalla simultaneously in offline mode so latency was generated only in the 5G environment.Moonlight’s on-screen display was recorded and parsed into CSV latency metrics.
  • Cloud Gaming: 24,06ms mean latency characterized srsRAN across configurations, with C2 on N310 achieving 23,54 ms mean latency and 5,31ms mean jitter.OAI showed unstable behavior, high jitter, and several outliers.
  • Cloud Gaming: 68,70% better mean latency made OAI C2 on B210 its best configuration relative to OAI C3 on N310.Frame drops remained close to zero despite high latency and jitter in some cases.

V. CONCLUSION

The conclusion combines qualitative platform assessment with preliminary performance analysis. OAI performs best for RRC Setup, while srsRAN is closest to theoretical data rates and consistently best for real application workloads.

  • V. CONCLUSION: OAI supports 120 kHz SCS in FR2, while srsRAN is notable for its well-structured documentation.The platforms share many similarities but differ in highlighted capabilities.
  • V. CONCLUSION: OAI achieved the best results for the RRC Setup procedure, although the overall performance of both platforms was similar.This conclusion follows the paper’s qualitative overview and preliminary performance analysis.
  • V. CONCLUSION: srsRAN reached up to 70% of TT downlink and about 30% of TT uplink across the evaluated SDRs and configurations.These results were closest to the theoretical values in the data-plane evaluation.
  • V. CONCLUSION: srsRAN consistently delivered the best results across all evaluated real application workloads.Future work will test higher bandwidths and evaluate deployment computational costs.
Loading 2608.25820v1…