IBM Storage supports NVIDIA DGX A100 AI infrastructure

By | 3 minute read | May 28, 2020

Artificial intelligence is driving innovation in all aspects of business. Customers can still struggle, however, with the building blocks of AI infrastructure, often resulting in the creation of multiple silos, or poor performance. This can lead to unnecessary complexity and increased costs. As companies strive to move quickly to deploy AI infrastructures to leverage vast amounts of data and drive business results throughout their organizations, they need to tap into the power of AI at scale.

IBM and NVIDIA solutions

Solving the problem of AI at scale is an increasingly important business need. Building on our existing collaboration, we are excited to unveil that IBM Storage for Data and AI now supports the recently announced NVIDIA DGXTM A100, which is designed for analytics and AI workloads.

IBM Storage for Data and AI recently updated the solutions blueprint with NVIDIA DGX2, leveraging the IBM ESS3000 NVMe all-flash system for high performance, IBM Cloud Object Storage with capacity efficiency and IBM Spectrum Discover to catalog and index the data.

We plan to have an updated blueprint leveraging IBM Storage for Data and AI for the NVIDIA DGX A100. With the fastest I/O architecture of any DGX system, NVIDIA DGX A100 is the foundational building block for large AI clusters such as ​NVIDIA DGX SuperPOD​, and the enterprise blueprint for scalable AI infrastructure. IBM brings together the infrastructure of both file and object storage with NVIDIA DGX A100 to create an end-to-end solution. It is integrated with the ability to catalog and discover (in real time) all the data for the NVIDIA AI solution from both IBM Cloud Object Storage and IBM Spectrum Scale storage.

Expanding our work together in HPC and AI

IBM Spectrum Scale is a leading choice for HPC and AI projects because it is readily deployed, high-performant and provides unique, intelligent scaling. It’s also the same technology behind the world’s most powerful and smartest supercomputer.

The US Department of Energy’s Oak Ridge National Laboratory has the IBMbuilt Summit supercomputer, currently used to identify and study 77 small-molecule drug compounds to fight against the SARS-CoV-2 coronavirus, responsible for COVID-19. Summit is powered by 9,216 IBM Power9 CPUs, over 27,000 NVIDIA V100 Tensor Core GPUs and 250 petabytes of IBM ESS Storage Systems using IBM Spectrum Scale software.

Fast results for research

Faster results are critical for AI workloads, and with more processing power we should continue to see even faster results. The NVIDIA DGX A100 features eight NVIDIA A100 Tensor Core GPUs, providing significant acceleration for all AI workloads. IBM plans to qualify AI architectures based on NVIDIA DGX A100, adding to its line of existing reference architectures with NVIDIA.

NVIDIA DGX A100

NVIDIA DGX A100

NVIDIA Magnum IO complements the high performance of IBM Spectrum Scale storage software and ESS storage solutions. IBM has been gaining a lot of traction with the latest ESS 3000 that utilizes containerized IBM Spectrum Scale for simplified management and upgrades. The efficient 2U building blocks start at under 25TB, with 40GB/sec performance and exabyte scalability. The IBM ESS 3000 is based on an NVMe-powered IBM Storage platform. It’s a preconfigured, tuned solution in a single box to accelerate time to value.

IBM’s NVMe-based ESS3000 driving 40GB/sec performance

IBM ESS 3000 and the entire IBM Storage for Data and AI infrastructure portfolio will continue to meet the growing performance requirements of AI workloads as they continue to evolve and drive faster business results.

IBM Storage for Data and AI simplifies your infrastructure with optimized efficiency to drive faster results that are massively scalable and globally available from edge to insight.

  • Simplify the journey to AI through a comprehensive data and AI portfolio
  • Simplify AI infrastructure for modern solutions from edge to core to cloud
  • Simplify multicloud and container connectivity with integration to IBM Cloud Pak for Data and IBM Storage Suite for IBM Cloud Paks

To find out more information on how IBM Storage simplifies your journey to AI with an integrated portfolio and provides solutions for your NVIDIA AI infrastructure, visit us here.

Most Popular Articles