AI, Security Front And Center As Pure Storage Looks To Simplify Management

The Santa Clara, Calif.-based Flash storage-focused technology developer is focusing on three major sets of capabilities to the Pure storage platform, said one of the firm’s executives.

Flash storage-focused technology developer Pure Storage Wednesday unveiled AI and security enhancements aimed at making it easier to manage and protect individual arrays and fleets of arrays.

With the news, Pure Storage is focusing on three major sets of capabilities to the Pure storage platform, said Prakash Darji, vice president and general manager for Santa Clara, Calif.-based Pure Storage’s Digital Experience business unit.

The first is the next-generation Pure Fusion hybrid storage cloud operating model, as well as an AI copilot. The second is enhancements to the company’s Evergreen//One storage-as-a-service subscription model with improved flexibility. The third, Darji told CRN, focuses on helping businesses expand from traditional workloads towards AI workloads as well as improving cyber resilience.

[Related: AI, Hyperscaler Cloud To Drive Pure Storage Business: CEO Charles Giancarlo]

“When we talk about Pure as a platform, that means one single operating system for scale-up and scale-out built on an evergreen architecture with simple management at any scale and powered by SLAs delivered via storage as a service,” he said.

Pure Storage bringing AI to its security and ransomware protection is the right direction because there are two things that customers are willing to spend money on—security and AI—even if it’s not in the budget, said Mark Gonzalez, regional vice president of sales at ePlus, a Herndon, Va.-based solution provider and long-time Pure Storage channel partner.

“AI isn't viewed as an expense,” he said. “It's viewed as a competitive weapon where people are willing to spend millions and millions and millions of dollars to try to gain a competitive advantage.”

Pure Storage is taking something—security—that has been traditionally considered sexy and turning it into something that is a competitive weapon that a customer can use to protect themselves, Gonzalez said.

“Everybody now assumes they will eventually be hit with a ransomware event,” he said. “It’s no longer just adding and adding and adding layers because you know, eventually it’s going to be penetrated. If you know that, then really the only weapon at your disposal is how quickly can you recover? … A condition of being able to stay in business is hardening the environment that you're in. If you’re a storage company like Pure and you can expand the touch points within your customer base, that's nothing but good.”

Pure Storage unveiled an expansion of its Pure Fusion that now allows policy-based fleet management at any scale, Darji said.

While Pure Storage’s technology is known for its management simplicity and plug and play capabilities, that’s at the array level, he said.

“How do we extend that to fleets [of arrays]?” he said. “We have some customers now with over 1,000 arrays, more data under management, more diverse types of workloads. So how do you manage at scale? We designed this next generation Fusion with the idea to make fleet management.”

Pure Fusion already automates storage environments with things like deciding where to place workloads, ensure that all of a configuration is compliant, manage governance and compliance drift, attach protection policies to workloads, and ensure those policies are consistently enforced, Darji said.

Pure Storage redesigned Pure Fusion so it is now part of its Purity software-defined technology, he said.

“It's embedded into Purity on every array,” he said. “When you log into the array itself, you can join a fleet. This extends the array management space to neighboring arrays to create this concept of a fleet. Let’s say you have 10 arrays. It doesn't matter which one you log into. You can provision, you can assign a policy such as deploying an Oracle workload where you say, ‘Deploy the primary on this X, deploy the secondary snapshots on C, set this retention policy, and deploy that across my fleet. It's available for provisioning for from one to thousands of arrays.”

The new fleet management capability is protocol agnostic, fully backwards compatible, and extensible to an entire fleet of arrays on an active cluster with certain distance restrictions, Darji said. Using asynchronous policies, the fleet could be located anywhere, he said.

Pure Storage is also taking advantage of the 25 petabytes of information it collects annually across its deployed arrays to start training a large language model to introduce a copilot for customers to interact and ask questions of their arrays, Darji said.

“They can ask things like, ‘How do I benchmark on my security across my fleet?’ or, “This Oracle volume has performance issues, what should I do?’” he said. “Our training is on our knowledge base as well as the data we've collected in Pure1. We're announcing this in private preview to customers at Pure//Accelerate because with LLMs, there’s a lot of reinforcement learning, a lot more training, to grow and scale. But we’re making it available for leading customers that want to go ahead and interact with the tech preview at Accelerate.”

Pure Storage has yet to determine general availability for its AI copilot, Darji said.

Also new from Pure Storage is Evergreen//One for AI, which Darji said is the first purpose-built AI storage-as-a-service for GPU performance aimed at ensuring customers can meet their SLAs (service level agreements) as their AI-related data grows.

AI is becoming more and more “mundane,” meaning that it touches everything, and when it comes to storage management, it makes total sense, ePlus’ Gonzalez said.

“In storage, it’s the mother lode because the whole basis of AI is data, right?” he said “This is this is a beautiful place for storage partners to really sell a lot of storage, especially with customers looking for simplicity at the same time as scalability. And Pure is a leader in terms of flash storage. Barbara Mandrell used to have a song that said, ‘I was country before country was cool.’ Pure Storage was flash before flash was cool. They've done a lot of things to get first-mover advantage. It’s a great company.”

With virtual machines, databases, or other traditional workloads, users typically scale performance to meet increased data requirements, Darji said. With AI, it is important to optimize utilization of GPUs, one of a data center’s most expensive assets.

“You need guaranteed provision throughput,” he said. “Whatever SLA you need on provisioned throughput, we will deploy whatever hardware needed under the Evergreen//One, and you can pay a marginal consumption rate for the data. It's a fairly new model to provide guaranteed provisioned bandwidth for GPU utilization optimization with marginal rates for data usage. We'll deploy or change whatever hardware needed to meet those SLAs.”

Extending SLAs to disaster recovery makes sense, Gonzalez said.

“Pure Storage has a very good offering in terms of ransomware protection, etc.,” he said. “They're very, very proud of their storage as a service because they tend to be one of the leaders in these types of things. They already offer the ability to scale to infinity by placing gear ahead of time at the customer's environment they can dial up or dial down so and then they have an SLA in terms of uptime with guaranteed financial penalties if they don't meet it. So this makes total sense. It's just an extension of something they've already been doing.”

Pure Storage is also working with Nvidia on certifying the jointly-developed Nvidia DGX SuperPOD AI data center infrastructure to be deployed on Ethernet networks, Darji said. The SuperPOD is typically used with InfiniBand, but the two vendors expect an Ethernet version available in the second half of 2024 depending on Nvidia’s timing, he said.

Pure Storage is also looking to secure application workspaces in Kubernetes containers in multi-tenant environment with new policy governance tools, Darji said.

“We took our Portworx storage container management software and linked it to secure multi-tenancy where you can create a secure application workspace for data agents top to bottom for LLMs. You can assign a policy for securing a management space for a data agent from your container and then use that to have a secure tenant guaranteed and isolated on your storage system.”

Cyber resiliency in AI environments is also a big focus of Pure Storage that goes beyond the typical data snapshots, safe mode, and data reduction-based anomaly detection, Darji said. With AI, however, attack vectors are getting a little bit more complicated, he said.

Pure Storage is responding with a new security assessment based on the aggregated intelligence from over 10,000 environments to score a company’s entire storage fleet and deliver best practices aligning with NIST 2.0 standards, Darji said. The new assessments also helps ensure regulatory compliance, fix potential security anomalies, and quickly restore operations after a security event.

“Our new AI copilot is also trained on it,” he said. “You can ask security questions around how to resolve things and benchmark against peers. Our enhanced anomaly detection, looking at performance, is new. And then our cyber recovery and resilience SLA is expanding our ransomware recovery SLA to all disasters, and adding this proactive element of remediating security issues found from our security assessment, proactively. And we're also introducing updates to our software patching where now you could do fleet-wide mass remediation.”

Finally, Pure Storage is reserve expansion, recommendation, and simulator, Darji said. This works not only with the company’s storage, but with storage from a hyperscale cloud provider like Amazon Web Services, he said.

This also includes a new site balance SLA in which a customer who signs up for the program can re-balance its workloads across on-premises and cloud infrastructures once a year with Pure Storage providing new hardware as needed, he said.