VAST Announces New GPU Reference Architecture with NVIDIA

VAST has announced their new Reference Architecture with NVIDIA for the DGX A100. This state of the art architecture incorporates the DGX A100 and VAST Lightspeed Universal Storage platform to deliver petabyte scale AI infrastructure solution for AI workloads that are GPU and storage intensive.

XENON is proud to be exclusive partner for VAST in Australia and New Zealand, and we have delivered multiple VAST systems since the relationship started in 2019. We can help you design an end to end infrastructure for your AI requirements using the proven VAST reference architectures.

From VASTdata.com:

New York – January 21, 2021 – VAST Data , the storage company breaking decades-old tradeoffs, today announced a new reference architecture based on NVIDIA DGX™ A100 systems and VAST Data’s Universal Storage platform. This reference architecture is designed to significantly increase storage performance for AI use cases such as large-scale training of conversational AI models and petabyte-scale data analytics. Jointly designed, built and tested by NVIDIA and VAST Data, the reference architecture eliminates the guesswork of building a solution from the ground up by providing enterprises with a turnkey petabyte-scale AI infrastructure solution that maximizes performance without adding needless cost and complexity.

Historically, enterprises have been forced to choose between two specific infrastructure configurations based on workload (i.e. GPU-intensive or storage-intensive), but as applications and data science teams’ requirements evolve, that infrastructure choice may limit and negatively impact performance. Built on VAST Data’s recently released LightSpeed platform and NVIDIA’s universal AI system, DGX A100, the architecture leverages VAST’s unique capabilities such as NFS-over-RDMA, NFS Multipath, and support for NVIDIA GPUDirect Storage as well as a converged fabric design. The joint reference architecture delivers more than 140GB/s of throughput for both GPU-intensive and storage-intensive AI workloads.

“For the first time, enterprises, and more importantly data science teams, are no longer constrained by the limitations of rigid infrastructure configurations,” said Jeff Denworth, Co-founder and CMO at VAST Data. “We’ve worked with NVIDIA on this new reference architecture, built on our LightSpeed platform, to provide customers a flexible, turnkey, petabyte-scale AI infrastructure solution and to remove the variables that have introduced compromise into storage environments for decades.”

“AI workloads require specialized infrastructure, which is why we’ve worked with VAST Data, a new member of the NVIDIA DGX POD ecosystem, to combine their storage expertise with our deep background in optimizing platforms for AI excellence,” said Tony Paikeday, Senior Director of AI Systems at NVIDIA. “This new reference architecture provides customers with a winning formula to achieve their AI success.”

For more information on VAST’s GPU Reference Architecture, including specific tests showcasing both infrastructure level microbenchmarks as well as benchmarks representative of real-world AI training workloads, please download the VAST Refence Architecture whitepaper now.

VAST and NVIDIA will also be participating in a webinar with IDC, “Smart Coupling of Next Generation AI Computing with Hyperscale Low-Cost Flash,” view this on-demand webinar now.

VAST Data AI-Ready Ecosystem

“Providing our customers with the right technologies to drive outcomes for their clients has remained an important focus of ours. We recognize that data-intensive AI will always be a big part of that strategy. With NVIDIA’s DGX A100 servers and VAST’s Universal Storage Platform we can provide a petabyte-scale AI infrastructure solution for GPU-intensive and storage-intensive AI workloads.” – Dragan Dimitrovici, CEO XENON Systems