How NVIDIA, VMware Are Bringing GPUs to the Masses

Last week virtualization giant VMware held its VMWorld 2019 user conference in San Francisco. Even the 23,000 or so attendees were treated to notable innovation the virtualization in the host company as well as its many partners. Among the more interesting statements that I believe flew under the radar was the combined NVIDIA–VMware initiative to bring virtual images processing unit technology (vGPU) to VMware’s vSphere and the VMware cloud on Amazon Web Services (AWS).

Virtual GPUs have been in use for some time but weren’t available to operate on virtual servers. Now businesses can conduct workloads, such as artificial intelligence and machine learning, using GPUs on VMware’s vSphere.

It Ought to step up and own GPU-accelerated servers
Historically, workloads that required GPUs had to operate on bare metal servers. This meant each info science group in a company had to buy its own hardware and incur that cost. Additionally, because these servers were only used for those GPU-accelerated workloads, they have been frequently secured, deployed and managed outside of IT’s control. Now that AI, machine learning and GPUs are moving somewhat mainstream, it is time for IT to step up and take ownership. The challenge is it doesn’t wish to accept the task of running dozens or hundreds of bare metal servers.

GPU sharing is your top use case vGPUs
The most obvious use case for vComputeServer is GPU sharing, where several virtual machines can share a GPU–similar to that which server virtualization did for CPUs. This should enable businesses to accelerate their information science, AI and ML initiatives because GPU-enabled digital servers can spin up, spin down or migrate like all other workloads. This will drive usage upward, enhance agility and help businesses save money.

This invention should also lead to businesses being able to run GPU-accelerated workloads in hybrid cloud environments. The virtualization capabilities combined with VMware’s vSAN, VeloCloud SD-WAN and NSX network virtualization produce a solid foundation for a migration into running virutual GPUs in an actual hybrid cloud.

Customers can continue to leverage vCenter
It’s important to comprehend that vComputeServer works with other VMware applications like vMotion, VMware Cloud and vCenter. The extended VMware service is important because this lets enterprises take GPU workloads into highly containerized surroundings. Additionally, VMware’s vCenter has become the de facto benchmark for data center management. At one time I thought Microsoft might struggle here, but VMware has won this particular war. Therefore it is logical for NVIDIA to enable its clients to deal with the vGPUs via vCenter.

NVIDIA vComputeServer also enables GPU aggregation
GPU sharing should be game changing for many businesses interested in AI / ML, which should be nearly every company today. But vCompute Server supports GPU aggregation, which enables a VM to get over one GPU, which is often a requirement for compute intensive workloads. VComputeServer supports multi-vGPU and peer reviewed computing. The distinction between the two is that with multi-vGPU, the GPUs can be distributed and not connected; with peer reviewed, the GPUs are linked using NVIDIA’s NVLink, making multiple GPUs seem like one, more powerful GPU.

A number of years back, using GPUs has been limited to a handful of niche workloads performed by specialized teams. The more data-driven businesses become, the further GPU-accelerated processes will play a key role in not only artificial intelligence but also day-to-day usable intelligence.

Together, VMware and NVIDIA have created a way for organizations to get started with AI, data sciences and machine learning without needing to break the bank.

Zeus Kerravala is an eWEEK frequent contributor and also the founder and principal analyst with ZK Research. He spent 10 years in Yankee Group and before this held a number of corporate IT positions.