VMware and Nvidia expanded their partnership to allow enterprises to deploy artificial intelligence (AI) workloads in Kubernetes containers.

The collaboration, which was announced at this week's VMworld 2021 event, pairs VMware’s vSphere for Tanzu Kubernetes platform with Nvidia’s recently announced AI Enterprise software suite. The latter, Nvidia claims, provides customers with everything they need to develop AI applications for health care, manufacturing, and financial services to name a few. The integration with Tanzu enables those workloads to be deployed as containers in a Kubernetes cluster running on existing data center infrastructure.

“This is one of the ways that we're helping bring AI in more quickly into the enterprise environment where we see the need for enterprise resiliency [and] quality of service,” Lee Caswell, VP of product marketing at VMware, said as part of a pre-show press conference. “All of these things are now important ways for us to make AI mainstream and business critical.”

The announcement makes good on a promise made at last year’s VMworld where the vendors announced plans to integrate Nvidia’s software capabilities with VMware’s vSphere, Cloud Foundation, and Tanzu platforms.

“Modern AI workloads can demand specialized infrastructure and software, creating complexity for IT teams working to support these advancing application requirements within enterprise data centers and hybrid clouds,” John Fanelli, VP of enterprise product management at Nvidia, explained in a blog post.

The vendors took the first step toward addressing this challenge this spring, announcing the aforementioned AI Enterprise software suite alongside an integration with VMware’s vSphere 7 hypervisor. The integration enabled AI workloads — which have traditionally been run on bare-metal servers — to run in virtual machines (VMs), with very low overhead.

According to Nvidia, the approach significantly reduced deployment times and enabled AI workloads to be spread across multiple nodes without a substantive performance hit. Or, for multiple smaller workloads to be run simultaneously on servers equipped with an Nvidia A100 GPU.

Today’s announcement builds on these capabilities, extending this functionality down to the container level. And similar to the earlier vSphere announcement, the vendors claim to be able to achieve near bare metal-like performance even in containerized workloads. According to Fanelli, a recent MLPerf benchmark submission by Dell Technologies achieved 94.4% to 100% of the performance of an equivalent bare metal environment.

Nvidia AI Enterprise is available for VMware vSphere today, and evaluation licenses are now available for vSphere for Tanzu.

Nvidia also announced a collaboration with Lenovo to promote the adoption of VMware’s Project Monterey, which entered early access in August.

Announced at VMworld 2020 after nearly three years of development, Project Monterey provides a consistent platform for distributed compute across multiple hardware accelerators using data processing units (DPUs), like those supplied by Intel and Nvidia.

“This is giving us a really interesting opportunity to go and help customers locate application and data processing at the right element of the computing infrastructure,” Caswell said.

In essence, the platform allows users to stitch VMs together from resources distributed across the data center rather than being limited to a single server.

“By putting ESXi managed through vCenter, we can start building a distributed virtual machine. To do that we need to have a control software that actually can manage aggregating and pooling of resources,” Paul Turner, VP of product management for VMware’s Cloud Platform Business Unit, said in an earlier interview.

Nvidia and Lenovo’s collaboration enables enterprises to trial Project Monterey remotely on VMware-enabled Lenovo servers, which have been equipped with Nvidia Mellanox’s BlueField-2 DPUs.

In other words, the program allows customers to test out the technology before committing to it at scale.