Meet AIRI: AI-Ready Infrastructure That Scales for Every Enterprise

NVIDIA and Pure Storage partner to simplify and accelerate the deployment of AI infrastructure at scale.
by Tony Paikeday

When tapping into the power of AI, enterprises soon find themselves thinking about how their data centers will need to evolve to embrace GPU-accelerated deep learning. And if their mission involves building AI-infused business applications, products or services, they’ll also find themselves thinking about scale. In which case, we’ve good news for them in a new offering called AIRI — AI-Ready Infrastructure — built on Pure Storage FlashBlade and NVIDIA DGX-1.

For many businesses, their datasets (and the opportunities buried within them) easily justify a significant investment in deep learning compute power. Autonomous driving software development is a great example of this. Fleets of hundreds of vehicles, each generating over a terabyte of data per hour, are causing developers to focus massive investment in GPU-powered deep learning.

Where some organizations might struggle to keep a single deep learning appliance busy, many others need dozens or even hundreds of nodes to shrink model training times from years to days.

The challenge lies in whether you’re ready to take on the task of developing and deploying an architecture that can deliver the scalable performance you need — and whether you have the resources to put it in place.

For organizations seeking an accelerated path to deploying an architecture that scales, Pure Storage and NVIDIA have introduced AIRI, a converged infrastructure offering for deep learning. AIRI extends the power of our NVIDIA DGX-1 AI supercomputer, combined with Pure Storage FlashBlade, and is backed by enterprise-grade support for the entire stack.

AIRI uses the NVIDIA DGX-1 integrated with Pure Storage FlashBlade as a modular building block for scaling AI capacity in any enterprise. With it, organizations can take advantage of the same design best practices that some of the world’s largest data centers have brought to bear as they race to tap insights from information faster than ever.

AIRI can help enterprises create a massive AI infrastructure — without a massive commitment of money, resources and time. Why we like it:

  • Its design was shaped by real-world deployments, with the NVIDIA DGX-1 and Pure Storage FlashBlade working in tandem to help eliminate bottlenecks that impede the ability to keep the GPUs inside the DGX-1 fed with data. Paige.AI is using the technology to accelerate its efforts to further clinical diagnosis and treatment in oncology using AI. Another customer, Element AI sees the opportunity for “shattering the barrier of infrastructure complexities” with AIRI, both for their business, as well as their customers’.
  • It’s been validated by a joint Pure Storage and NVIDIA team and backed by a published reference architecture. With this documentation, customers can have peace of mind regarding performance expectations from deploying an AIRI-based infrastructure in their data centers.
  • It offers a simple, modularized means by which customers can scale up without incurring a lot of design guesswork or time spent building and tuning systems. Enterprises can reap the know-how of those who’ve done this already, in a ready-to-deploy architecture available through, and supported by, our NVIDIA deep learning partners in North America: ePlus, FusionStorm, Groupware, Trace3 and WWT; and, in the Europe, PNY.

To learn more, join our webinar to get a deep dive on AIRI and explore how it might offer you a new way to unleash AI at scale in your enterprise.