VMware, Nvidia team up to extend IT infrastructure for AI

VMware and Nvidia have partnered to create an enterprise-wide platform for AI to tie together the data center, cloud and edge computing.

The design goal of the new collaboration is to merge Nvidia's NGC GPU software hub with VMware's vSphere, Cloud Foundation and Tanzu.

The two companies believe the new platform will serve to accelerate the adoption of AI and better enable corporate users to extend their current infrastructure for AI, as well as manage all their applications, through a single set of operations. This gives users the flexibility of deploying AI-capable infrastructure wherever the data resides.

Pat Gelsinger

"This [partnership] helps us make the GPU a first-class compute citizen through our network fabric and VMware virtualization layer," said Pat Gelsinger, VMware's CEO, in a session at the VMworld 2020 digital conference. "This is critical to making [GPUs] enterprise-available to all VMware users and not some specialized infrastructure tucked in the corner of the data center. It will be available to every virtual infrastructure admin."

Some believe VMware's decision to form a partnership with Nvidia, which has relationships with many top-tier cloud and hardware suppliers, is a necessary one given VMware's market share in both the cloud and AI markets dominated by AWS, Microsoft and Google.

"[VMware] can see what is rapidly happening with the leading competitors and they realize they have to do something because they are a lower-tier player in [cloud and AI] markets," said Frank Dzubeck, president of Communications Network Architects, Inc., an IT consultancy in Washington, D.C. "This is why they go to a chipmaker that has been really aggressive in these areas."

[VMware] can see what is rapidly happening with the leading competitors and they realize they have to do something because they are a lower tier player in [cloud and AI] markets.

Patrick Moorhead, president and principal analyst with Moor Insights & Strategy, agreed the deal should benefit corporate users of both companies.

"This announcement means there is a full stack integration of Nvidia's AI stack, which makes it easier for enterprises to leverage and share Nvidia GPUs," Moorhead said.

The UCSF Center for intelligent Imaging, a user of both VMware and Nvidia products, said it plans to integrate the two ecosystems as part of its ongoing development of AI and analysis tools in medical imaging. The center uses the Nvidia Clara healthcare application framework, along with the VMware Cloud Foundation, for a number of mission critical applications.

AI has been instrumental in detecting diseases in large imaging studies, and does so much faster than the human eye can, said Christopher Hess, chairman of Radiology and Biomedical Imaging at UCSF. Soon, AI will be able to offer the most accurate diagnoses and treatment for patients. He said the AI application frameworks along with VMware Cloud foundation will enable the center to expand its work in AI by using a common infrastructure for research and development.

Jensen Huang

"This is the moment in history where advanced breakthroughs in medical research are critical," said Jensen Huang, Nvidia's CEO. "As people search for vaccines, some of the most confidential patient information, which is the crown jewels for big pharma, is also in play, so there are great concerns over security. And this is where we can help with AI breakthroughs."

Both Gelsinger and Huang see opportunities to deliver a new architecture for the hybrid cloud that will assist users in evolving their infrastructure and operations to deliver a new security model that offloads hypervisor, networking, security and storage capabilities from the CPU to the data processing unit (DPU). This proposed architecture will also be able to extend the VMware Cloud Foundation operating to bare metal servers.

The new architecture will also serve as the foundation for VMware's Project Monterey, which was previewed during VMworld. By tying the NVIDIA Bluefield-2 DPU with VMware's Cloud Foundation, users can further speed the performance of both existing and next generation applications. It will also accelerate those applications beyond AI to other enterprise-class apps, and provide an added layer capable of offloading data center services from the CPU to programmable DPUs.