NVIDIA, AI, and Pavilion - Shattering Expectations for the Future of Technology

On Monday, April 12, 2021 at GTC21, NVIDIA Founder and CEO Jensen Huang gave an impressive keynote where he laid out the future of computing and how technology will shape every aspect of the human condition.

The path forward is driven by accelerated computing, AI software that writes software, the data center as the new unit of computing, a new industrial revolution powered by AI on 5G, and autonomous systems.

All of this is built on the foundation of NVIDIA computing technology, including announcements on GPUs, DPUs, and CPUs, as well as an AI powered virtual world known as the NVIDIA Omniverse, and AI on 5G to accelerate the Enterprise and Industrial Edge.

This is a vision for what the data centric world can do for humanity.

All of this is dependent upon the combination of three pillars:

  1. Parallelized data processing powered by NVIDIA solutions,
  2. High speed, low latency networking based on Mellanox technology,
  3. A performant, dense, scalable, and flexible data storage platform.

The first two pillars are based on NVIDIA technology.

The third pillar, requires an unmatched data storage solution that can deliver unlimited scale with high performance, ultra-low latency, and the flexibility to support any data type.


The NVIDIA technologies can only achieve their potential if they use a storage solution that can quickly deliver data to the processing resources.

This is what led NVIDIA CTO Michael Kagan to say at SuperComputing20 that, “Storage access is yet another bottleneck that needs to be resolved and Magnum IO and GPUDirect storage seamlessly utilize capabilities of advanced networking and enables data transfers at line rate between storage server and the GPU computer”. Mr Kagan went on to say that Pavilion was able to deliver “Performance improvement for file storage of almost 4x by using GPUDirect Storage”.

There are many choices available to customers considering a storage solution. Though, in this new world, leveraging a parallel storage architecture that demonstrates monumental performance differences is required to align with the vision of NVIDIA.

In a recent Blocks and Files article, Storage Product Comparisons on the Rack, Chris Mellor normalizes vendor claims which enables organizations to have a realistic understanding of which storage architecture is best suited for environments that require significant throughput, low latency, density, and flexibility.

The Pavilion HyperParallel Data Platform delivers unprecedented data storage performance, density, scalability and flexibility while offering customers unrivaled data storage choice and control across block, file, and object data.

To learn more about how Pavilion and NVIDIA solutions can unlock the power of data for your organization, watch this HPCwire webinar featuring NVIDIA and Pavilion.

You can also learn more about Pavilion partners OmniSci, Graphistry and the broader NVIDIA ecosystem by attending the following GTC sessions:

High Performance Storage - How a Multi-Controller Storage Architecture Shatters Expectations for Modern Applications 

Visually Investigating Patterns in Logs at Scale with Graphistry, RAPIDS, and Pavilion 

How High Performance NVMe-oF Storage Accelerates CPU- and GPU-Powered Virtualized Environments

How to Respond to Rapidly Scaling Geospatial Intelligence with OmniSci and Pavilion