How High Performance NVMe-oF Storage Accelerates CPU & GPU-Powered Virtualized Environments Demonstrated by Pavilion

VMware, Pavilion and NVIDIA power GPUs on vSphere 7 RoCE - On-Demand GTC Session # SS33026

There are clear ESXi benefits around security, simplification and proactive support from VMware. Pavilion doubles-down on these with Universally Unmatched Storage to consolidate islands of aging SAN or NAS storage, deploy NoSQL scale-out databases on vSphere 7 with RoCE, and run AI/ML workloads in virtualized environments with the most performant, dense, scalable and flexible storage platform in the universe. We have clear quantifiable benefits.

Saving ½ the budget
One of Pavilion’s newest customers took advantage of vSphere 7 and RoCE to bring new products to market in record time while doubling their VM/VSI density and cutting storage costs in half, while leaving fibre channel SANs in the rear-view mirror.

Virtualized NoSQL is faster and cheaper
Another account has gone a step further. Rather than run the GreenPlum database on bare metal resources, they virtualized the entire NoSQL scale-out database on the Pavilion HyperParallel Data Platform™, gaining 25% improvement in IOPs while reducing the overall footprint by 50%. Yes, better and faster NoSQL virtualized. Others are taking note.

AI/ML/DL just got better on VMware
In another project, Pavilion is teaming with a customer to implement vSphere 7 Update 2 with containers to decouple physical GPU resources from their servers and shatter expectations for AI/ML/DL pipelines.

Wait, what?
AI/ML/DL in a virtualized environment???

On March 9, NVIDIA VMware announced a certified suite of AI tools, the NVIDIA AI Enterprise Software Suite allowing customers to give workloads direct access to Nvidia’s CUDA applications, AI frameworks, pre-trained models and software development kits. Users can now share resources and improve GPU utilization while managing the entire environment under ESXi. The new software is certified to run Nvidia’s A100 Tensor Core GPUs.

What used to require bare metal servers with direct-attached storage can now use VMs or containers for AI training and inference as well as data analytics and machine learning workloads.  With NVMe-oF and RoCE, workloads will be performant and easier to manage.    According to Justin Boitano, general manager of NVIDIA’s enterprise and edge computing unit, “the performance of vSphere is virtually indistinguishable from bare metal.”

Pavilion demonstrated how high performance NVMe-oF storage accelerates CPU & GPU-powered virtualized environments at the NVIDIA GTC21 conference in a session with Pavilion Chief Field Technology Officer, Costa Hasapopoulos and Director of Product and Solutions Marketing, Walter Hinton. 

If you want to learn more about how the Pavilion is shattering expectations for GPU-based workloads, check out these on-demand sessions and recent blogs:

How to respond to rapidly scaling Geospatial Intelligence with OmniSci and Pavilion

How a multi-controller storage architecture shatters expectations for modern applications 

Visually investigating patterns in logs at scale with Graphistry, RAPIDS, and Pavilion