Dell Upgrades Storage Offerings As Businesses Shift To AI-First Operating Model

The total volume of stored data worldwide is expected to hit 175 zetabytes by 2025, with 90 percent of that taking the form of unstructured data – including images, video and sensor data – and that is where Dell is focusing many of its advancements in optimizing storage for AI.

As more businesses see the competitive need to shift data into AI-driven business processes Dell Technologies is ramping up its AI-Ready Data Platform offering, which combines infrastructure and system performance upgrades, to speed up how quickly customers can train large language models.

“At the heart of this data platform is our ability to connect,” said Martin Glynn, Dell’s senior director of product, in a media briefing. “First we’re providing customers a modern data lakehouse. We’re continuing our efforts there. That provides a single point of data access. We then extend that into an open ecosystem where we work with lots of partners to bring the best toolsets into the customer environment for the purposes of data analytics, building, ML workflows, and also out to the cloud ecosystem.”

Ninety-two percent of businesses and IT executives agree that their organization needs to shift to an AI-first operating model withing the next 12 months to stay competitive, Glynn said. That requires extensive compute, networking and storage capabilities.

Glynn said that with Dell’s market leadership position in servers and storage systems, the company is ready to help with validated designs built on AI-optimized infrastructure.

“The vast majority are planning to shift to an AI-first operating model. The way they’re going to do that is to use their own data as the differentiator for their AI operations. We’re all familiar with the continued explosive growth in data…”

Dell PowerScale storage has been optimized to provide up to 80 percent gains in performance to handle the demands of AI work when using the PowerScale OneFS operating system, Glynn said. Additionally, PowerScale’s All Flash Nodes improve streaming reads by 2x and streaming writes by 2.2x.

The “reads” feed GPUs for faster model training while the “writes” play a critical role in speeding up the fine-tuning of models. The upgrades improve density-per-rack and lower total cost of ownership, Glynn said.

“We’re really excited to get these new performance enhancements and hardware into our customers’ hands because this performance has such a dramatic impact on the new kinds of activities that they’re taking to build their AI and generative AI operations,” Glynn said.

PowerScale will also be the first ethernet storage system validated on NVIDIA DGX SuperPOD. This leads to increased flexibility and speed for easer AI storage, Glynn said.

“PowerScale will be the first storage product to be validated to be part of an Nvidia DGX SuperPOD with enternet,” Glynn said. “Up until now all the storage that has been attached there has been inifiniband. Nvidia’s excited, we’re excited, with the increased flexibility and opportunity that Ethernet attached here will offer customers who are looking to deploy DGX Superpod.”

Dell Apex File Storage In Azure

Dell said it is also extending support for its Dell APEX File Storage to the Microsoft Azure cloud platform. Glynn said the company is working with Microsoft to take advantage of OpenAI and Vision Services.

“So the same set of tools and capabilities that Microsoft has allowed cloud customers to take advantage of, now they can take advantage of using PowerScale inside Azure,” Glynn said. “And they can seamlessly move data between their on-prem activities and their cloud activities, to take advantage of those capabilities.”

Modern Data Lakehouse

For Dell the ideal spot for AI workloads is what it calls a Modern Data Lakehouse where large unstructured pools of data reside. Greg Findlen, senior vice president of data management at Dell Technologies said enterprises often think they must first consolidate all their data in a single location before they can let their data scientists use it.

“A real critical part of what we’re focused on here is how do we help customers embrace the data gravity and access the data where it is?” Findlen said.

Dell’s Modern Data Lakehouse is multi-cloud by design, so the customer can store, process, and secure data on premise, in the cloud and at the edge. It de-couples computing and storage and reduces the need for data movement. All of this is done to remove complexity and speed up data discovery, querying and processing.

“How do we help them unlock the potential for that data without having to move it all into one place, whether it’s one place in the cloud or one place on premise,” said Greg Findlen, senior vice president of data management at Dell Technologies. “A big part of this vision here is delivering an integrated turnkey solution for our customers.”

The next-generation, all-flash Dell PowerScale systems and new smart scale-out feature will be globally available in the first half of 2024, as will be Dell APEX for File Storage for Microsoft Azure. The Dell lakehouse solution also will be available in the first half of 2024 incorporating Dell’s PowerEdge servers, PowerScale and – through a partnership with Starburst Data – that company's data lake analytics software. The system’s open architecture lets operators use Iceberg and Delta Lake to meet customer trends in open source software and open data, Findlen said.

“The key objective here is to make sure we can help customers discover, and integrate, and process their data across the organization,” he said.