Most data centers weren’t designed with the explosive growth and unique demands of AI in mind.
But NVIDIA DGX POD offers a blueprint for every data center architect who wants to optimize their compute, storage and network infrastructure to meet the rising tide of AI-infused applications.
And the industry’s data center leaders — including Arista, Cisco, DDN, Dell EMC, IBM Storage, Mellanox, NetApp and Pure Storage — have rallied around DGX POD with offerings based on their differentiated technologies.
Boost the ROI of Your AI
Machine learning and deep learning used to be science projects in the labs of academia. Now they’re mainstream capabilities that are growing the top line for businesses and helping drive down costs.
But unlocking that return on investment means having an AI-ready data center that can scale. And many organizations don’t for the simple reason that AI workloads consume resources unlike any other tier-one task.
At NVIDIA, we perform some of our most important, barrier-breaking work on a massive infrastructure, called SATURNV, that we built on NVIDIA DGX systems. In fact, we recently set six AI performance records in the MLPerf benchmark suite using a configuration based on the NVIDIA DGX-2. We put the system together in just under three weeks, using the design principles and architecture on which SATURNV is based.
Build Your Own SATURNV
Arista, Cisco, DDN, Dell EMC, IBM Storage, Mellanox, NetApp and Pure Storage — they all deliver turnkey infrastructure solutions using NVIDIA DGX. These integrated systems give customers a proven, validated approach that aligns with their choice of data center technology partner. They enable every enterprise to build its own SATURNV — sized to meet their needs, of course.
AI Infrastructure News at GTC
At GTC this week, our ecosystem has big news to share regarding DGX POD:
- NetApp ONTAP AI is now available with NVIDIA DGX-2, in addition to its current DGX-1 offering. With systems based on the DGX-2 and the revolutionary NVSwitch GPU networking fabric, NetApp is enabling organizations to build infrastructure optimized for unprecedented levels of model and data parallelism. By combining cloud-connected flash with the world’s first single-node 2 petaflops system, NetApp ONTAP AI accelerates outcomes at any scale.
- Pure Storage is announcing Hyper-scale AIRI, which takes AI infrastructure deployment to a new definition of scale and is built on Pure Storage Flashblade, NVIDIA DGX-1 and NVIDIA DGX-2. The success of AIRI is based on a scalable formula that can start as small as two nodes of DGX-1 systems and can now grow seamlessly to 64 racks of optimized AI infrastructure.
- The most recent addition to the DGX POD ecosystem is the Dell EMC Isilon with NVIDIA DGX-1. It’s now commercially available as an integrated, end-to-end, turnkey AI solution offered through leading channel partners in North America to start, with a global rollout coming soon.
- Red Hat is bringing together the trusted Red Hat Enterprise Linux with NVIDIA DGX systems. By adding the certification of DGX-2 on Red Hat Enterprise Linux, organizations don’t have to choose between developer productivity and performance vs. IT operations compatibility. They can have both, now certified with the 2 petaflops NVIDIA DGX-2 system.
DGX-Ready: Putting an AI-Ready Data Center Within Reach
Building on DGX POD is the formula for getting AI at scale. And there’s good news for those who don’t have the data center or cap-ex to make it happen soon enough. A growing roster of data center colocation service providers know exactly how to get you from plan to production.
Each is certified for their ability to stand up and host DGX infrastructure. They can bridge your facilities and capital budget gap with an attractive, outsourced op-ex model for running AI infrastructure built on DGX POD.