China Mobile Uses NVIDIA GPU Servers to Deliver AI over 5G at the Edge.
When natural disasters strike, responders race against time to deploy critical resources and save lives.
Fanned by strong winds, a forest fire raged through a remote southwestern corner of China’s Sichuan province in early April. Drones equipped with high-resolution cameras and infrared detection technology were dispatched to the mountainous terrain, from where they transmitted footage of the leaping flames over 5G networks to emergency dispatch headquarters.
Responders, rather than waiting for drones to return to start processing the data, could immediately begin parsing the video with AI image algorithms running on NVIDIA GPUs, helping them better understand the crisis and concentrate rescue efforts.
This groundbreaking work was led by the China Mobile Chengdu Institute of Research and Development a research division of China Mobile, the world’s largest mobile network operator — using advanced 5G technology, AI and the China Mobile Link-Cloud platform for drones.
The company, which has nearly a billion customers, is accelerating natural disaster response, improving emergency medical services and providing new education tools with NVIDIA GPUs connected to next-gen 5G mobile networks.
For example, a joint rescue team from China Mobile’s research division and Sichuan Provincial People’s Hospital in June used ambulances equipped with 5G terminals to remotely diagnose patients at the scene of a 6.0 earthquake. First responders in the emergency vehicles conducted tests like ECG monitoring or ultrasounds, using low-latency 5G networks for real-time video consultations with doctors at the hospital.
There, physicians could use GPU-accelerated medical imaging AI to diagnose and provide temporary treatment instructions until the patients were transferred to the hospital for surgical treatment.
Elsewhere, high-bandwidth 5G towers help address educational inequality between urban and rural areas by connecting multiple classrooms through virtual reality. China Mobile has connected a classroom from a rural primary school in Sichuan with students in Chengdu, the province’s capital. To do so, they used VR headsets, NVIDIA GPUs and an integration of the NVIDIA CloudXR software development kit — which delivers low-latency AR/VR streaming over 5G networks — with an application for remote synchronization of classrooms.
This initiative could help thousands of schools in far-flung regions participate in the same real-time, interactive learning experiences as more resource-rich schools.
Future deployments of these pilot projects will shift computational processing from data centers to GPUs at the edge, whether embedded in drones and ambulances or in full racks of edge servers.
Deploying 5G to 600 Million Users
With 10x lower latency and 1,000x the bandwidth of existing networks, 5G makes data-intensive mobile computing applications such as 4K video and VR possible at the edge for the first time. It also enables the deployment of complex AI models for inference at the edge.
China Mobile, a leader in 5G deployment, has to date installed 50,000 5G stations across 50 cities in China. The country is projected to have 600 million 5G users by 2025.
The company is a member of the Open Data Center Committee, a nonprofit consortium formed by the country’s leading technology providers and telecom giants. One of the committee’s initiatives is the Open Telecom IT Infrastructure (OTII) project, an effort to standardize server solutions for 5G mobile edge computing.
NVIDIA EGX servers developed by data center systems provider Inspur and edge computing manufacturer ADLINK will be the first GPU hardware to be incorporated under the OTII standard.
Powered by NVIDIA T4 and NVIDIA Quadro RTX GPUs, respectively, servers like these can be used at the edge to accelerate critical AI applications using 5G networks. An end-to-end software development kit compatible with Chinese technical requirements for mobile edge computing has also been developed to facilitate GPU adoption.
The NVIDIA EGX edge computing platform consists of a cloud-native software stack and edge servers optimized to run the stack. EGX systems vary from NVIDIA Jetson-powered edge devices to NGC-Ready for Edge validated servers. With NVIDIA EGX, system administrators can easily set up a fleet of edge servers remotely and securely for faster, easier deployment.
Inspur, H3C and Lenovo are among the dozens of manufacturers worldwide offering EGX systems today.