Pure Storage has announced a suite of new AI features, including the introduction of a new ‘copilot’ feature and enhanced support for GPUs.
The firm unveiled these new features as part of its flagship Pure//Accelerate 2024 event in Las Vegas as it looks to aid businesses in AI deployment and AI management of storage assets.
On the management level, Pure Storage is looking to give its customers a greater level of control through its new ‘Copilot for Storage,’ a platform which will allow users to monitor and manage their storage environments via natural language inputs.
The company said the new AI tool leverages data insights from “tens of thousands” of customers to assist storage teams in complex tasks surrounding performance and management.
Copilots are increasingly all the rage as generative AI continues to gather momentum, though Pure thinks its own offering is unique and described it as an “industry-first”.
At a press conference ahead of the firm’s annual event, Shawn Hansen, GM at Pure Storage, said the new tool “fundamentally changes the way that you simplify managing storage”.
“Instead of having to search through 10 different screens to be able to find something, you can ask a question in your own language,” Hansen said.
Copilot will simplify how users find solutions to complex storage problems, Hansen added. For example, instead of having to call customer service to deal with the “rocket science” that is a latency problem across many different arrays, users can just ask Pure’s copilot.
“AI does an amazing job of looking across billions of rows of information … we will automate some of the most difficult needles in a haystack to find,” Hansen said.
The AI tool will have additional advantages in cyber security by providing a greater level of visibility to CISOs into benchmarking their security posture against other Pure Storage customers, the company added.
Pure Storage lifts the lid on new STaaS offerings
The firm also introduced ‘Evergreen//One™ for AI,’ a storage-as-a-service (STaaS) focused on streamlining AI deployment and offering “guaranteed” storage performance for GPUs to support training and inference.
The new Evergreen//One platform also offers the ability to purchase “dynamically” based on performance and throughput needs, offering customers an increased level of flexibility in terms of AI storage requirements.
While Pure Storage already offers a range of Evergreen platforms, as well as an existing Evergreen//One platform, this new announcement highlights the firm's shift to ensuring AI provisions.
Pure also rolled out ‘AI-Powered Anomaly Detection Enhancement’ which allows users to discover various cyber security threats such as ransomware attacks or unusual activity.
The firm said the models used to underpin this platform analyze customer environments with “historical data for anomalous patterns based on heuristics of performance as well as user context on how storage is used”.
Pure Storage is going at AI from different angles
There are “multiple components” to consider in terms of the market for AI, according to CEO Charlie Giancarilo.
Speaking to ITPro at the pre-event press conference, Giancarlo said the company aims to support customers through a multi-pronged approach centered around bolstering the underlying infrastructure needed to drive AI development.
Data center capacity - or the lack of it - has been a key hurdle identified by enterprises seeking to explore the use of generative AI over the last 18 months.
The GPU cloud market, for example, looks for “raw horsepower and performance and scalability,” Giancarlo said, and Pure has a specific set of programs for these sorts of users.
“That's our Flash Blade product, and we continue to develop that set of capabilities,” Giancarlo said.
That’s a “unique market,” though, with a comparably smaller number of customers, Giancarlo added. Most enterprises are looking at smaller environments to develop AI and, for these markets, Pure is looking at “two main activities”.
“One is to be able to deploy ready made, highly configured, inference type environments,” Giancarlo said. Complete with vector databases, these environments would be focused on simple deployment models for AI.
The second aim for Pure is to help businesses deal with the issue of data, which is often spread out across enterprise environments. As arrays are often “highly tuned” to the applications and stacks they’re tied to, businesses can end up copying data into new arrays dedicated to AI.
“We're saying that's a really silly and old fashioned thing to do,” Giancarlo said.
Pure’s view, Giancarlo added, is that businesses should ensure that “all” of their storage is accessible to AI as they look to upgrade their storage environments. This also involves moving to flash storage to enhance performance.
“We now can sell all-flash systems at the same price as existing disk-based storage, and when we do so the performance is at least at least five times higher, if not more” he said.
