NVIDIA uses graphics processors and artificial intelligence to bring 5G connectivity to the mobile edge. At the Mobile World Congress Americas in Los Angeles, California, Nvidia CEO Jensen Huang announced that the company has partnered with Ericsson and Red Hat to help mobile network operators accelerate 5G deployments and better utilize the bandwidth of their applications of interest. Connect games, virtual reality and augmented reality, autonomous driving and smart home to 5G networks
Partnerships address some of the shortcomings of current mobile network architectures. For example, the 4G LTE network is configured to be optimized for maximum capacity with dedicated hardware to provide voice and data access, said Justin Boitana, Enterprise and Edge Computing Manager at Phone and Briefing prior to Hwang's MWC keynote. Explained. In urban areas, this means that when a user leaves at night, the carrier cannot redeploy their existing bandwidth, which wastes network capacity when the user leaves at night.
When 5G is released, it will have 1,000 times the bandwidth and one tenth of latency. However, one of the advantages of this standard is that the infrastructure introduces a new architecture called network licensing or network slicing. This allows service providers to dynamically provide unique services taken from common data center pools and share infrastructure across multiple cell towers.
“What our customers want is to be able to flexibly scale the right services based on consumer data needs in a software-defined way at every point in time. “Now operators can collaborate with Ericsson or create a 5G radio access network or RAN, the largest network execution software, to run many services on a shared infrastructure. By using off-the-shelf hardware, operators can now deliver a lot of services at the mobile edge with 5G and see maximum investment utilization for networks that can offer a variety of new services, including AR, VR, as well as games for consumers. . AI for companies that don't want to run the EGX server itself. ”
The virtualized network will run on the wireless infrastructure closest to the consumer and is ideal for delivering cutting edge AI services, the company claimed.
Nvidia used a natural language processing model to compare the performance of GPU-accelerated EGX hardware with existing CPU-based servers, claiming that a single EGX rack provides equivalent performance to about 60 dual-socket Intel Xeon processors. NVIDIA's Aerial Network SDK is currently available for testing by developers, and the software kit includes an optimized packet processing and signal processing pipeline that leverages the power of the GPU.
NVIDIA and Ericsson's partnership will inject the latter expertise into 5G wireless access networks as leaders in high performance computing to deliver 5G supercomputing and artificial intelligence on a single platform. Nvidia is also working with Red Hat to transform the carrier network into a software-defined network so that more devices can connect and achieve this. The Red Hat partnership will bring carrier-grade Kubernetes to the mobile industry.
"The collaboration announced by NVIDIA founder and CEO Jensen Huang can help carriers transition to 5G networks that can run a variety of software-defined edge workloads," he said. "This study will focus on 5G Radio Access Networks (RANs) to further improve the communication of AI-based applications."
Editors' Picks