Unified data platforms reshape AI storage

Unified data platforms reshape the basics of business storage as the adoption of AI is accelerating, forcing companies to revise aging architectures and to rethink the way the data is managed, set up and consumed.

Faced with the spread of assembly data and requests for workloads accelerated by the GPU, organizations turn to storage solutions that unify block workloads, files and objects under a rationalized framework. The change is urgent: the inherited systems cannot keep the rhythm of the AI ​​-oriented environments which require flexibility, centralized control and the ability to transparently manage unstructured data sets through the deployments of hybrid clouds. At the heart of this transformation is an increasing thrust to simplify complexity, reduce operational costs and unlock the full potential of large scale AI, according to Rob Strechay (Photo, right), main analyst of Thecube Research.

“Growth was phenomenal with Alletra’s deputy and released 18 to 24 months,” said Strechay. “Really what they do with it … Unified data services bringing together complete management cycle of the equipment and making this so simple, as well as giving this cloud consumption model under Greenlake.”

Strechay spoke with another analyst Savannah Peterson (left), because they provided an analysis of their recent interview with the interview with Patrick Osborne for the Interview series “Cloud ai Journey with Hpe”During an exclusive broadcast on TheCube, the live studio from Siliconangle Media. They discussed how unified data platforms allow the scalability of AI, the growing role of hybrid deployments and the growing importance of metadata and inference in modern storage strategies. (* Disclosure below.)

Unified data platforms take care of hybrid scalability and flexibility

The integration of AI into modern IT environments highlights the need for unified data formulas which can offer flexibility without sacrificing performance. Organizations are moving away from fragmentary storage solutions in favor of workload systems that rationalize management and simplify infrastructure, according to Strechay.

“This is why people like the cloud and really like to buy in this methodology,” he said. “I think the adoption of customers … like Patrick [Osborne] Said, they voted with their dollars, which was great to see. »»

This change is also linked to improving the experience of the developer and customers, Peterson said. Unified data services provide not only technical scalability, but also create smoother and more intuitive user routes.

“You highlighted the three things that stood out in this particular part of the [Osborne] Maintenance also, “noted Peterson.” It is a single value proposal that is really aware of the experience of these developers and their customers. “”

The discussion also covered how Hybrid and on -site deployments Check for renewed attention, especially since AI models require high -speed access to large data sets, a large part of which always lies in business walls. The problems of confidentiality, regulation and performance of data maintain significant workloads of AI on site, according to Strechay.

“I think a lot also aimed in large part to be hybrid,” he said. “Hybrid and on -site deployments, we see … 85% of the data used for AI is in fact on site these days.”

Metadata, inference and simplification of the IA storage complexity

Another point to remember key has revolved around metadata and how companies prioritize not only the data itself, but the information on this data. As the workloads focused on AI are on the scale, metadata management becomes a critical differentiation, allowing faster recovery, smarter organization and more efficient treatment.

“A quote that I want to call,” said Peterson. “”[Osborne] Said something that is really important that I think that sometimes people are missing when they start thinking about their strategy around it. He said: “that data on data is almost as important or more important than the data itself”.

Metadata has become vital in modern storage systems. AI inference is closer to the data itself, allowing companies to reduce latency and avoid the movement of useless data – an approach that Strechay fully takes care of. He pointed out on the transition to the integration of inference and generation with recovery directly in storage platforms to improve efficiency and performance.

“What [Osborne’s] There are the two different types of metadata that people use to build these AI systems, “he said.” When you start looking at it, he was really aware. Corporate data is essential. It is the heart of the company; It is the heart of organizations and it is their intellectual property. »»

Although AI often feels as magic, it is the result of planning, development and engineering extended behind the scenes, according to Peterson. Adopting simplicity in the very complex data infrastructure today requires significant efforts and expertise.

“I think it’s a key … that making something simple that is as complex as these storage and data platforms is not at all easy,” added Strechay. “It’s about making it a simple customer experience.”

The conversation should continue to Hpe Discover.

Here is the complete video interview of Thecube, which is part of the cover of Siliconangle and Thecube of the Interview series “Cloud ai Journey with Hpe”::

https://www.youtube.com/watch?v=ozvd8iybd50

(* Disclosure: Thecube is a paid media partner for the interview series “Cloud AI Journey With Hpe”. Nei Hewlett Packard Enterprise Co., The sponsor of the coverage of Thecube events, nor other sponsors have editorial control over the content on TheCube or Siliconangle.)

Photo: Siliconangle