NetApp and Nvidia are taking the latest steps to give enterprises the same capabilities to leverage artificial intelligence technologies that currently are primarily the domain of major cloud-based companies and the largest corporations.
The two companies are bringing together Nvidia’s GPU-based DGX supercomputers and NetApp’s AFF A800 cloud-enabled all-flash storage to create what officials are calling the ONTAP AI architecture to help enterprises better manage the massive amounts of data being generated not only in the data center but in the cloud and at the network edge. It’s also designed in a modular fashion that will make implementation and scaling easier, addressing some of the key challenges enterprises face when trying to deploy such AI techniques as machine learning, deep learning and natural language processing into their environments.
“For all the focus these days on AI, it’s largely just the world’s largest hyperscalers that have the chops to roll out predictable, scalable deep learning across their organizations,” Jim McHugh, vice president and general manager of deep learning systems at Nvidia, wrote in a post on the company blog. “Their vast budgets and in-house expertise have been required to design systems with the right balance of compute, storage and networking to deliver powerful AI services across a broad base of users.”
Enterprises face a range of challenges when trying to bring AI into their operations, from leadership issues and a lack of skills to the complexity of the technologies and a shortage of resources. A growing number of established vendors such as IBM, Microsoft, Hewlett Packard Enterprise and Cray and startups like H2O are offering products designed to address those challenges. Nvidia has made AI a target growth area for the past several years, including through such GPU platforms as the DGX systems.
AI is expected to have a significant impact on businesses as they look to gain greater insights from the large amounts of data being generated, particularly as more intelligent devices—from the smallest sensors to large industrial systems—become connected. Gartner analysts in April said that the business value worldwide derived from AI will hit $1.2 trillion this year, a 70 percent increase over 2017, and will jump to $3.9 trillion in 2022. The value will come from an improved customer experience, new revenue from new products and more sales, and reduced costs.
"One of the biggest aggregate sources for AI-enhanced products and services acquired by enterprises between 2017 and 2022 will be niche solutions that address one need very well,” John-David Lovelock, research vice president at Gartner, said in a statement. “Business executives will drive investment in these products, sourced from thousands of narrowly focused, specialist suppliers with specific AI-enhanced applications."
Gartner noted that improvements in deep neural networks and compute power were key drivers of AI. Another is the amount and speed of data being created. According to Octavian Tanase, senior vice president of ONTAP at NetApp, a key challenge for companies is being able to gather and manage the huge amounts of data being generated. In a post on the company blog, Tanase pointed to smart asthma inhalers as an example of using AI at scale. The devices bring in data not only about the patient but also about such areas as the weather, air quality and pollen counts, all of which can play a factor in triggering an attack.
“Data flows from thousands of devices at the edge,” he wrote. “That data is combined with outside data sets during training in a core on-premises data center with GPU acceleration. The resulting inference model is deployed in the cloud to analyze new data points and identify and act on trigger events.”
Any bottleneck with the data not only increases costs and wastes time, but “in the smart inhaler example and many other use cases, bottlenecks also put outcomes at risk,” Tanase wrote. “It does a patient little good to find out about a trigger event after they’ve had an attack.”
The NetApp-Nvidia ONTAP AI offering is designed to help enterprises more easily scale their AI deployments by reducing design complexity and offering a deep learning architecture that can scale as the data grows, according to officials. It also includes redundant storage, network and server connections to reduce bottlenecks.
The connectivity is delivered via NetApp’s Data Fabric, which can be used in the cloud or on-premises to tie together the various sources of data across multiple clouds, including public clouds, private clouds in data centers and hybrid cloud environments. The company’s AFF A800 can scale from two to 24 nodes. Nvidia’s DGX-1 supercomputer includes eight Tesla V100 GPUs, uses the company’s NVLink interconnect and offers 1 petaflop of compute power.