OpenAirInterface Setup

5G Stack Overview

Fig. 20Overview of the deployed 5G end-to-end stack. Figure fromOpenAirInterface.

The 5G stack is deployed as Docker containers.Fig. 20 shows the system architecture with network interfaces and IP addresses.

The Sionna Research Kit builds GPU-accelerated (CUDA) images for:

  • oai-gnb-cuda - gNodeB with GPU acceleration

  • oai-nr-ue-cuda - 5G NR UE with GPU acceleration

  • oai-flexric - FlexRIC for O-RAN support

The 5G Core Network usespre-built images from Docker Hub.

For further details, see theOpenAirInterface Documentation.

Note

The following steps can also be executed via:

makesionna-rk

Or individually with the following commands:

# Pull, patch and build OAI containers./scripts/quickstart-oai.sh# Generate configuration files./scripts/generate-configs.sh# Build plugin components (TensorRT engines, etc.)./plugins/common/build_all_plugins.sh--host./plugins/common/build_all_plugins.sh--container

Manual Build Steps

For development or debugging, the individual steps are documented below.

Clone the OAI repository:

gitclone--recurse-submodules--branch2025.w34https://gitlab.eurecom.fr/oai/openairinterface5g.gitext/openairinterface5g

Apply the Sionna Research Kit patches:

cdext/openairinterface5ggitapply--index<../../patches/openairinterface5g.patch

These patches enable GPU acceleration and contai the plugin infrastructure for the Sionna Research Kit.

Run the docker builds individually:

cdext/openairinterface5g# Base imagedockerbuild--targetran-base-cuda--tagran-base-cuda:latest\--filedocker/Dockerfile.base.ubuntu.cuda.# Build image (run from sionna-rk root to include plugins)cd../..dockerbuild--targetran-build-cuda--tagran-build-cuda:latest\--fileext/openairinterface5g/docker/Dockerfile.build.ubuntu.cuda.cdext/openairinterface5g# gNodeBdockerbuild--targetoai-gnb-cuda--tagoai-gnb-cuda:latest\--filedocker/Dockerfile.gNB.ubuntu.cuda.# UEdockerbuild--targetoai-nr-ue-cuda--tagoai-nr-ue-cuda:latest\--filedocker/Dockerfile.nrUE.ubuntu.cuda.# FlexRICdockerbuild--targetoai-flexric-fixed--tagoai-flexric:latest\--filedocker/Dockerfile.flexric.ubuntu.

Note thatran-build-cuda must be built from the Sionna Research Kit root directory so that plugins are included in the build context.

Check that all images were built successfully:

dockerimagesREPOSITORYTAGSIZEran-base-cudalatest17.2GBran-build-cudalatest21.9GBoai-gnb-cudalatest16.4GBoai-nr-ue-cudalatest8.1GBoai-flexriclatest1.1GB

Build Plugins

Some plugins must be built on the specific platform. For example, the TensorRT engines are target specific:

./plugins/common/build_all_plugins.sh--host./plugins/common/build_all_plugins.sh--container