NetApp and Nvidia announce Ontap AI, an AI data platform for enterprise

Synthetic intelligence isn’t nearly algorithms. The information on which they’re skilled is extra vital than the fashions themselves, some would argue, which is among the causes IDC predicts that greater than 44 zettabytes of digital information might be created by 2020. Fortunately coinciding with the rise of “huge information” is a continued decline in cloud storage pricing, motivated partly by cheaper media prices, higher administration instruments, and improvements in object storage.

However not all cloud storage suppliers are created equal. Some lack the fine-grain administration instruments required to collate, course of, and switch AI mannequin information rapidly and effectively. And never all enterprises have storage stacks optimized for information science workflows.

Fortunately, there’s an answer: Ontap AI, an “AI confirmed structure” Nvidia and information storage firm NetApp at present collectively introduced. Powered by Nvidia’s DGX supercomputers and NetApp’s AFF A800 cloud-connected flash storage, it’s designed to assist organizations obtain “edge to core to cloud” management over their information by delivering unprecedented entry and efficiency, stated Octavian Tanase, senior vp at NetApp.

“Our distinctive imaginative and prescient of a knowledge pipeline [affords] simplicity of deployment,” he advised VentureBeat in a telephone interview. “Persons are on the lookout for scale — they wish to begin small and develop. On the finish of the day, we would like clients to have the ability to handle information throughout the sting … correlate information units, construct giant information lakes, [and] finally make quicker choices and higher choices [about] information.”

The connective tissue that ties it collectively is NetApp’s Knowledge Cloth, a SaaS and on-premise answer that unifies information sources throughout clouds — particularly, clouds in information facilities, public cloud choices from service suppliers, and hybrid non-public/public clouds — and gives lightning-fast entry to information no matter its format or bodily location.

On the core of Ontap AI is Nvidia’s DGX-2, a second-generation deep learning-optimized server that sports activities 32GB of GPU VRAM, 16 V100s GPUs, and a brand new high-performance interconnect referred to as NVSwitch. A single DGX-2 rack delivers as much as two petaflops of compute energy, Nvidia claims, and might practice neural machine translation fashions corresponding to FairSeq in as little as a day and a half.

NetApp’s aforementioned AFF A800s, in the meantime, boast equally spectacular efficiency: sub-200 microsecond latency and throughput of as much as 300GB/s in a 24-node cluster.

“On the earth of AI, information integration is crucial,” Jim McHugh, vp and common supervisor at Nvidia, stated in an interview. “What’s actually required for GPU AI coaching is sort of completely different than conventional purposes. The aim is to make it as painless as doable for information scientists, and as painless as doable for the individuals who construct out infrastructure, too.”

One in every of Ontap AI’s first practitioners is Cambridge Consultants, a U.Okay.-based engineering consulting agency. It utilized it in a well being care vertical, the place it’s leveraging it to construct methods that consider drug remedies and their impacts on affected person outcomes. It’s additionally used it to create Vincent, a deep studying program designed to learn to paint like a human,

Different Ontap AI launch companions embody IAS, Groupware Expertise, ePlus, and WWT.

“Creating disruptive AI know-how and turning this into breakthrough services for our clients is an important requirement throughout many markets we work in,” Monty Barlow, head of synthetic intelligence at Cambridge Consultants, stated.  “[It’s] simplifying and accelerating the information pipeline for deep studying.”

Nvidia’s highly effective {hardware} platform is a giant get for NetApp, which reported internet income of $5.9 billion in Might for its fiscal 2018. It’s the one storage vendor partnered with the three largest public cloud suppliers — Amazon Net Providers, Microsoft Azure, and Google Cloud.

Show More

Leave a Reply