As an increasing number of applications and related complexity put unprecedented demands on computing infrastructures, our customers are realizing that the future of computing needs to be more heterogeneous in nature; a single technology, no matter how innovative, cannot address all the requirements of modern, let alone future computing.  

Keeping a large number of systems running smoothly is a hard problem, and special devices are often required to help run well-balanced and optimized data center infrastructure. These devices, called data processing units (DPUs), employ an easily programmable multi-core CPU, a state-of-the-art network interface and a powerful set of networking, storage and security accelerators that can be programmed to perform multiple software-defined, hardware-accelerated functions. 

Red Hat and NVIDIA's collaboration

Fitting inside nearly any traditional data center server, the NVIDIA BlueField-2 DPU is supported by Red Hat Enterprise Linux (RHEL), the world’s leading enterprise Linux platform that underpins Red Hat’s industry-leading software, enriching existing capabilities by offloading specialized tasks such as network address translation, data compression, and deep packet inspection, among others. To learn more about Red Hat’s view of emerging datacenter solutions, including DPUs, read Chris Wright's post about driving computing innovation.

These new DPU devices not only can aid the datacenter services running on them, but the actual servers that customers use to run their business applications and other critical software. The goal is to make the installation and configuration of server hardware, including GPUs, more streamlined across the entire enterprise software stack. 

That is why we are continuing our joint work to certify the latest NVIDIA DGX A100 systems with RHEL 7 and 8 while offering simplified NVIDIA GPU driver packaging in RHEL. General availability of these precompiled driver packages will provide customers with an easy-to-deploy foundation for GPU enablement across the hybrid cloud, from bare metal servers and virtual machines to Linux containers and cloud environments. 

Red Hat is also in the process of joining NVIDIA’s DGX-Ready Software program to offer Red Hat OpenShift, the industry’s leading enterprise Kubernetes platform, as a foundation for ecosystem innovation. Customers will have access to proven, tested, enterprise-grade software solutions certified with OpenShift on clusters of NVIDIA DGX systems. This can help simplify the deployment, management, and scaling of AI infrastructure, while ecosystem partners can tap OpenShift to develop and deliver solutions to customers in a more scalable and repeatable way.

We’re also working with NVIDIA to make it easier for our joint customers to collaborate on and run advanced graphics and computing workflows from any location, especially in light of dynamically changing global conditions. This requires virtualization and abstraction layers in Red Hat Virtualization and Red Hat OpenStack, to be aware of the presence of GPUs -- and, of course, able to access them. We are supporting these deployment scenarios along with multi-instance GPU capabilities and GPU virtualization on NVIDIA A100 GPUs to enable a new range of AI-accelerated workloads that can run on top of vGPUs and Virtual Compute Server software from NVIDIA.

To learn more about our ongoing collaboration with NVIDIA, please attend our sessions at NVIDIA's GTC 2020 virtual event and hear from Red Hat’s experts first hand. Learn more about Red Hat's presence at NVIDIA GTC 2020. You could also see what’s happening in the OpenShift community by attending customer and partner presentations at the OpenShift Commons Gathering. This gathering on AI and Machine Learning is co-located with GTC 2020.