A GPU Hyperconverged Platform for 5G vRAN and Multi - Access Edge Computing

In this paper we present the NVIDIA hyper-converged platform supporting 5G connectivity and Mobile Edge Computing (MEC). 5G connectivity is realized with our Aerial [1] GPU-based cloud native 5G gNB. We introduce AI-on-5G on a converged accelerator to showcase our innovation in being able to host Aerial vRAN baseband processing, AI/ML training and inference, data analytics and other workloads. In other words, a data center at the edge that is provisioned with 5G connectivity as a service. We describe 3 uses-cases that highlight how existing NVIDIA AI/ML development frameworks, together with Aerial, can be leveraged to bring Industry 4.0 to reality. As an open platform Aerial is positioned to be industry transformational by providing researchers with a platform for next generation wireless and AI research. Aerial seeds the research ecosystem with a first-class out-of-the-box (OOB) experience with a standards compliant 5G NR PHY. Researchers can run the supplied 3GPP compliant test vectors, and perform over-the-air experiments, using standard servers equipped with a GPU-based PCIe card. The PHY code base can be tailored to support research that combines AI/ML with 5G wireless.