At the Computex event in Taiwan, NVIDIA unveiled EGX, a multi-cloud and AI-enabled edge computing platform for enterprises.
NVIDIA EGX is a unified edge computing stack that can span from the tiny Jetson Nano to a full rack of T4 servers. Customers can start small with EGX and gradually scale to support full-blown GPUs. NVIDIA is optimizing the software stack to power devices such as drones to dedicated servers that can handle AI inferencing at scale.
NVIDIA Edge Stack is an optimized platform powered by NVIDIA drivers, a CUDA Kubernetes plugin, a CUDA container runtime, CUDA-X libraries and containerized AI frameworks and applications such as TensorRT, TensorRT Inference Server and DeepStream SDK. Customers can run a variety of AI workloads including computer vision, natural language processing and speech synthesis.
Red Hat and NVIDIA have partnered to deliver EGX on OpenShift, an enterprise container orchestration platform based on Kubernetes. Red Hat OpenShift, which can be deployed on-premises or in the public cloud, provides a consistent development and deployment experience to businesses. The combination of OpenShift and CUDA makes it possible to train machine learning models in the cloud and deploy them seamlessly at the edge for inferencing.
NVIDIA is also partnering with Cisco to bring security, networking and storage technologies to the edge. The stack also has components such as smart network interface cards and switches from Mellanox, a company that NVIDIA acquired earlier this year.
The EGX software stack is tightly integrated with Azure IoT and AWS IoT, making it a multi-cloud IoT and edge computing platform. It also runs Azure IoT Edge and AWS IoT Greengrass – two major edge runtimes – offered by Microsoft and Amazon. This integration makes EGX one of the first multi-cloud IoT and AI platforms.
At the time of the launch, EGX servers are available from global enterprise computing providers including ATOS, Cisco, Dell EMC, Fujitsu, Hewlett Packard Enterprise, Inspur and Lenovo. Enterprise IoT Gateway providers such as Abaco, Acer, ADLINK, Advantech, ASRock Rack, ASUS, AverMedia and Supermicro have also announced the availability of EGX stack on their appliances.
NVIDIA is pitching EGX to enterprises from the smart city, healthcare, retail and manufacturing verticals. NVIDIA claims that BMW Group Logistics, a supply chain management entity, integrated AI directly to the edge of its logistics processes to handle complex logistics with real-time efficiency.
With EGX, NVIDIA got many things right. From partnering with Red Hat on OpenShift integration to packaging Cisco’s networking gear to a tighter integration with Azure and AWS, it pushed all the right buttons.
NVIDIA is firing on all cylinders to become a leader in AI infrastructure. From the public cloud to data center to the edge, CUDA and TensorRT are fast becoming the de facto standard for training and inferencing machine learning models. EGX is a step towards gaining the market share in the edge computing segment.