VMworld 2018: NVIDIA GPUs Build Strong Digital Foundations

by Anne Hecht

To build strong digital foundations, IT teams need the right tools.

Last week’s VMworld 2018 keynote, presented by VMware CEO Pat Gelsinger and CTO Ray O’Farrell, highlighted the role of NVIDIA’s virtualization technology in that effort.

They showed how NVIDIA GPUs accelerate organizations, enabling them to run deep learning, inferencing, training and HPC workloads on the same infrastructure that runs VDI workloads by repurposing the host after hours with vMotion support.

The keynote demonstrated image recognition using GPU-accelerated inferencing, running on a Quadro virtual Data Center Workstation (Quadro vDWS), Tesla V100 GPUs and VMware vSphere. The inferencing workloads doing image recognition on a land-use dataset had been previously trained on a deep neural network.

A single Tesla V100 running NVIDIA TensorRT can perform this common inferencing task 100x faster than a system without GPUs. The tech preview of vMotion technology support for NVIDIA GPUs is available on vSphere 6.7, update 1, with general availability coming soon.

Here’s what analysts are saying:

  • “Additionally, vSphere 6.7 Update 1 will increase support for intelligent workloads by introducing vMotion and snapshot capabilities for NVIDIA vGPU-powered VMs. This will enable admins to migrate vGPU-powered VMs to other compatible hosts while performing maintenance operations to completely eliminate any disruption to end-users and their applications.” — Computer Tech Review
  • “VMware has also thrown in additional support for ‘intelligent workloads’ thanks to new vMotion and snapshot capabilities for NVIDIA Corp.’s virtual graphics processing unit chips.” — SiliconAngle
  • “Besides the new interface, vSphere 6.7 Update 1 contains a vCenter Server Converge Tool and additional vMotion and snapshot capabilities for NVIDIA vGPUs …” — Tech Target

IDG Network World listed NVIDIA vGPU software support for VMware vMotion as a “hot product” at VMworld 2018. For users who want a turnkey solution for their software-defined data center, Dell EMC VxRail appliances was also listed. VxRail offers V series, which is a VDI-optimized, graphics-ready appliance with support for use cases such as high-end 2D/3D visualization.

On the VMworld show floor, real-time simulation on ANSYS Discovery Live was demonstrated on NVIDIA Quadro vDWS and Tesla P40 GPUs. A Windows 10 VDI environment showed a native, desktop-like experience running graphics-intensive websites on 4K resolution displays with low CPU utilization using NVIDIA GRID vPC and Tesla M10 GPUs. With the ability to support up to 32 knowledge workers (1GB profile), the Tesla M10 delivers the performance and user density required by the modern digital workplace.

vMotion of GPU-accelerated VMs was also demonstrated – the entire process of migrating a live VM was as simple as right-clicking on a VM to suspend, and then right-clicking to migrate it to a compatible host. The entire process took only a few seconds.

GPU Virtualization Sessions

If you missed the show, check out the replays of the sessions below from VMware and NVIDIA product leaders who highlighted the power of NVIDIA vGPU solutions in a VMware VDI environment:

Learn more about how NVIDIA vGPUs can help accelerate your VMware VDI environment.

Learn more about NVIDIA vGPU solutions by following @NVIDIAVirt.