Nvidia AI Enterprise Now Available on VMware vSphere Servers

Nvidia AI Enterprise, new software that lets partners build and run artificial intelligence workloads on VMware clusters, is now available. The provider of chipsets and software that accelerates compute and graphics on Tuesday released the new tools and modeling frameworks.

Introduced earlier this year, Nvidia AI Enterprise is available from channel partners and from key infrastructure vendors offering Nvidia-Certified Systems. It is the culmination of a partnership Nvidia and VMware announced last year seeking to bring AI to a broader set of enterprise customers. Nvidia AI Enterprise runs on industry-standard servers with vSphere 7.2, the latest release of VMware’s hypervisor platform.

The software paves the way for deploying AI-based infrastructure on-premises at scale, said Manuvir Das, Nvidia’s head of enterprise computing.

Nvidia’s Manuvir Das

Nvidia AI Enterprise “is specifically optimized to run on VMware so that customers can expect the same performance running virtualized in a shared VMware environment [as] running bare metal on servers today,” Das said during a briefing with press and analysts. “And it’s certified to run on a variety of hardware systems.”

Among those launching Nvidia-Certified Systems include Atos, Dell Technologies, Gigabyte, Hewlett Packard Enterprise, Inspur, Lenovo and Supermicro. Besides running Nvidia AI Enterprise, the servers are available with various Nvidia GPUs. This enables different scale-out and performance options, including its A100, A30, A40, A10 and T4.

The new Nvidia-Certified AI systems are mainstream servers that are accelerated with the GPUs, Das explained. VMware’s vSphere with Nvidia AI Enterprise allows multiple applications to share one system for traditional line-of-business applications as well as modern solutions such as business process automation and image analysis using AI.

“We rely on VMware vSphere, the de facto operating system on the data center to do management and orchestration,” Das said.

Domino Data Lab Partnership

Nvidia also announced that it has partnered with Domino Data Lab, which has agreed to validate the Domino Enterprise MLOps Platform with Nvidia AI Enterprise. Domino Enterprise MLOps provides machine learning and AI workflow beginning with the acquisition of data to production deployment, Das explained.

The MLOps platform provides life cycle management of software. This ensures that organizations know where specific AI and machine learning models originated. It also verifies that they have gone through testing and are compatible with the rest of the infrastructure.

Inside Nvidia AI Enterprise

Das explained that Nvidia AI Enterprise brings together the three key properties of building an AI infrastructure at scale. It consists of Nvidia’s Rapids SDK, which accelerates data science capabilities on its GPUs.

Nvidia uses AI frameworks including TensorFlow and PyTorch, and its existing training frameworks. Those retarget to the GPUs “so they can run much faster than before and at much greater scale,” Das said. Also, software called NvidiaTriton, which is used to run the inference models for the trained AI models.

“It’s a three-step process,” Das said.

First Certified HCI: Dell EMC VxRail

In addition to the Nvidia-Certified Systems that are servers, Dell Technologies announced the first certified hyperconverged infrastructure (HCI) solution certified with the new Nvidia AI Enterprise software.

Dell EMC VxRail is an automated system tested and validated with Nvidia AI Enterprise workloads, according to the announcement posted by Nancy Hurley, Dell Technologies’ senior manager for CI/HCI product marketing.

Dell’s Nancy Hurley

“VxRail HCI System Software delivers end-to-end automation and [life cycle management], and native integration with existing VMware management tools helps reduce the complexity of adopting Nvidia AI Enterprise in VMware environments,” Hurley noted. “Customers adopting Nvidia AI Enterprise can seamlessly manage their infrastructure and application from a single console so that AI resources can be managed more efficiently and effectively, saving time and money.”