VMware and NVIDIA Partnership Accelerates AI From On-Prem To The Cloud

The News: NVIDIA and VMware today at VMworld announced their intent to deliver accelerated GPU services for VMware Cloud on AWS to power modern enterprise applications, including AI, machine learning and data analytics workflows. These services will enable customers to seamlessly migrate VMware vSphere-based applications and containers to the cloud, unchanged, where they can be modernized to take advantage of high-performance computing, machine learning, data analytics and video processing applications. Read the full press release from NVIDIA.

Analyst Take: This week at VMworld, I expect the big theme to continue to be multi-cloud and Hybrid IT and for VMware it is all about building solutions that create the greatest level of seamlessness between on-prem and cloud. This has been the driving factor of vSphere as IT has sought for ways to build applications and workloads that can work on-prem or in the cloud consistently for both administrators and users.

To date, AI Training can be accomplished effectively on-prem or in the Cloud, but there has been some limitations on leveraging the benefits of hybrid IT to modernize applications using GPUs. The partnership between VMware, NVIDIA and AWS appears to be an enabling force that will help the enterprise maximize its GPU resources as well as create a truly dynamic hybrid environment where AI related workloads

A Closer Look At The Benefits:

Based upon my review of the announcements, the partnership zeros in on five specific areas of benefit for users.

• Greater Workload Portability: Single click portability of workloads using NVIDIA vCompute and GPUs on VMware HCX. This will give customers more choice and flexibility to execute training and inference in the cloud or on-premises while offering zero downtime.
• Leading Enterprise Cloud With AWS Integration: With the ability to automatically scale VMware Cloud on AWS clusters accelerated by NVIDIA T4, administrators will rapidly be able to attend to the needs of data scientist deploying the appropriate resources.  I would love to see a similar offering for Azure, but at this point the deeper integration is limited to AWS. I’ll keep everyone posted.
• Accelerated computing for modern applications: NVIDIA T4 GPUs feature Tensor Cores for acceleration of deep learning inference workflows. When these are combined with vComputeServer software for GPU virtualization businesses have the flexibility to run GPU-accelerated workloads like AI, machine learning and data analytics in virtualization environments for improved security, utilization and manageability. This configuration also allows GPUs to be shared among different training initiatives, so a single GPU may be able to work on several concurrent training sets much like how this is done with CPU in virtualized environments.
• Aministrator Friendly With Consistent Control Plane: With VMware Cloud on AWS, organizations can establish consistent infrastructure and consistent operations across the hybrid cloud, leveraging VMware industry-standard vSphere, vSAN and NSX as a foundation for modernizing business-critical applications. IT operators will be able to manage GPU-accelerated workloads within vCenter, right alongside GPU-accelerated workloads running on vSphere on-premises. This is really important as a big challenge for hybrid IT has been developing consistency for administrators to be able to operate on-prem architecture and their cloud in a similar fashion. This solution is designed to scale an organizations hybrid IT efficiency by making the workload portability seamless regardless of whether it is being moved from the cloud or to the cloud.
• Seamless, end-to-end data science and analytics pipeline: Users will benefit greatly by having access to NVIDIA’s technology specifically NVIDIA RAPIDS, a collection of NVIDIA GPU acceleration libraries for data science including deep learning, machine learning and data analytics.

Performance Impact and Tradeoff?

Yes, virtualizing the GPUs may have a small impact on GPU performance, but that is usually around 5% or less. I would expect the small degradation in performance to be a well-worth it trade off for the enhanced UX and other benefits of a seamless virtualized environment for proliferating enterprise data science efforts.

How Will Enterprises Consume (Read: Pay) 

The licensing will be on a per GPU basis. I can’t see a more effective or appropriate way to charge for these services.

In Short, It’s Important and Well Timed

As Hybrid IT continues to proliferate, it is becoming increasingly important that all workloads, including AI related are able to benefit from the experience that hyperscale cloud and virtualization have created. This partnership, between NVIDIA and VMware leveraging the power of AWS should enable greater efficiency of GPUs while offering a consistent control plane to administrators.

Futurum Research provides industry research and analysis. These columns are for educational purposes only and should not be considered in any way investment advice.

Read more analysis from Futurum Research:

Huawei Ascend AI Processors Show Its Ambition Despite Tensions

Qualcomm Wins Partial Stay In FTC Ruling, Overturn Likely To Follow

VMware Posts Earnings Just Ahead of VMWorld: Wins Keep Coming

Image: Flickr

The original version of this article was first published on Futurum Research.