The NextG-Lab testbed is a state-of-the-art experimental platform designed to advance research in 5G and 6G networking, edge computing, and AI-driven wireless communication. It follows the O-RAN architectural style and principles, offering a flexible and scalable environment for doing research and deployment of advanced end-to-end networks. In particular, it brings flexibility in investigating advanced radio access technologies, Multi Access Edge Computing (MEC)-enabled applications, and end-to-end softwarized network functions. Its architecture integrates a high-performance switching backbone, powerful gNodeBs with GPU acceleration, software-defined radio units, and versatile user equipment, enabling researchers to experiment with the network from an end-to-end perspective, i.e., from the User Equipment (UE) over network radio, transport, core, to the edge and cloud.
Starting from the user side, the testbed includes three Zotac ZBOX MI668 mini-PCs that act as UEs. Each UE is powered by a 12-core Intel i7-1360P CPU, 32 GB of DDR5 RAM, and 1 TB of NVMe storage. These devices are equipped with Quectel RM500Q 5G modems, enabling native 5G New Radio connectivity, and are also paired with USRP B210 SDRs for more advanced experiments that require custom PHY/MAC protocol development. This combination of commercial 5G modems and SDR-based flexibility allows the testbed to represent heterogeneous device scenarios, ranging from commercial-grade connectivity to highly experimental air interfaces.
To enable radio capabilities, the testbed integrates three USRP N310 Software-Defined Radios (SDRs), which serve as flexible remote units. These SDRs are capable of operating across a wide range of frequencies and bandwidths, making them ideal for prototyping and evaluating next-generation technologies. The gNodeB layer is implemented on five Dell Precision 5860 workstations, each powered by an Intel Xeon W5-2555X processor with 14 cores and 28 threads, operating at up to 4.8 GHz. Each node is equipped with 128 GB of RAM, a 1 TB NVMe SSD, and an NVIDIA RTX 4000 Ada GPU with 20 GB of VRAM. These workstations are further enhanced with four 10GigE interfaces, ensuring multi-path and high-bandwidth connectivity. The GPUs allow the gNodeBs to support real-time AI, and predictive resource allocation, while also enabling the execution of complex edge intelligence functions (Enricher).
On the transport side, a Cisco C9300X-48TX-A 10GigE switch, acts as the central backbone interconnecting all components. The switch supports multi-rate operation (1/2.5/5/10 GigE), ensuring high-throughput and low-latency connectivity between gNodeBs, SDRs, UEs, and external servers. This robust networking fabric allows the testbed to scale efficiently and to support demanding experimental setups.
The NextG-Lab integrates virtualized servers to host MEC and cloud-native applications. These servers, support virtual machines and containers orchestrated through platforms such as Kubernetes. This extends the testbed’s capability into service-based architectures, enabling experiments in NFV, network slicing, and Zero-Touch Service Management (ZSM). It also allows researchers to deploy edge applications, digital twins, and AI-driven orchestration mechanisms directly into the testbed environment.
Together, these components create a powerful and versatile testbed that is flexible and fully equipped to address the challenges of next-generation networking. The NextG-Lab supports end-to-end experimentation across the communication and computation continuum. Its modular and extensible architecture ensures that it can evolve alongside emerging technologies, making it a key tool for innovation in the path toward 6G.
In addition to its radio, compute, and user equipment components, the NextG-Lab testbed integrates a flexible and fully functional 5G Core network that enables end-to-end experimentation. The Core deployment is based on open-source frameworks, primarily Open5GS and OpenAirInterface (OAI), ensuring interoperability and alignment with ongoing academic and industry research. The testbed can also accommodate alternative implementations such as Free5GC, making it highly adaptable to evolving standards and experimental needs.
The 5G Core provides support for Standalone (SA) operation, allowing the testbed to run as a self-contained system that integrates both RAN and Core capabilities. This setup enables researchers to validate advanced 5G features, including network slicing, multi-tenancy, and service-based architectures, in both controlled lab environments and outdoor test scenarios.
The Core runs on high-performance computing resources, leveraging RTX 3060 GPUs for AI/ML integration, Intel processors (such as the i7-11700K with 8 cores and 16 threads), and up to 128 GB of RAM depending on deployment. To further extend computing capabilities, the NextG-Lab incorporates Proxmox-enabled Supermicro servers equipped with dual Intel Xeon CPUs and high-capacity memory, which support virtualized environments for deploying network functions, MEC workloads, and AI/ML services. These servers also integrate USRP X310 SDRs over 10GbE passthrough links, enabling seamless coupling of core functions with advanced PHY-layer experimentation.
Through this design, the Core not only supports the standard 3GPP functions, such as AMF, SMF, UPF, and PCF, but also provides a foundation for experimenting with programmable control loops and AI-driven orchestration, as well as network exposure principles through Open APIs such as those from the Camara project. Integration with FlexRIC, an open-source Radio Intelligent Controller (RIC), extends programmability further into the RAN domain, allowing researchers to study RIC-driven network optimization and O-RAN-based control.