Dell debuts new PowerScale storage systems and AI-focused software upgrades

Dell Technologies Inc. today debuted new additions to its PowerScale series of storage systems along with an update to the OneFS software platform that powers the product line.

The updates are designed to help customers run artificial intelligence models more efficiently. Dell is also rolling out enhancements to other product lines, including the PowerEdge server portfolio, that likewise focus mainly on supporting AI use cases.

PowerScale is a collection of NAS, or network-attached storage, systems optimized to hold unstructured data. The systems are powered by an operating system called OneFS. It mitigates the impact of hardware failures, detects ransomware and automates certain maintenance tasks for administrators.

Dell is adding new all-flash systems to the PowerScale lineup that it says can carry out streaming data read and write operations up to twice as fast as earlier hardware. The new systems are based on the latest machines in the company’s PowerEdge server portfolio. Dell says a new “smart scale-out capability” will enable companies to make more efficient use of the graphics cards attached to the servers in their PowerScale environments, which should improve AI workloads’ performance.

The OneFS operating system that ships with PowerScale appliances is also receiving an AI-focused upgrade. According to Dell, the new release of the software will allow customers to more quickly train AI models, fine-tune them and perform inference, or the task of running neural networks in production.

“Storage performance is a critical factor for successful AI and generative AI outcomes,” said Arthur Lewis, president of Dell’s infrastructure solutions group. “Customers are relying on us to continually push the boundaries of storage innovation, including removing data access bottlenecks that limit the throughput and scalability of compute-intensive applications.”

To expand the range of AI projects that PowerScale can support, Dell is validating the product portfolio for use in DGX SuperPOD environments. DGX SuperPOD is an offering from Nvidia that combines data storage hardware with the chipmaker’s GPU-equipped DGX appliances. Those appliances each include eight A100 or H100 graphics cards along with multiple switches, a large amount of memory and other supporting components.

Dell expects that PowerScale will be the first Ethernet-based storage system series to earn a DGX SuperPOD certification. PowerScale systems use Ethernet-based network links to move data to and from customers’ DGX appliances. Competing storage appliances rely on other networking technologies, namely Nvidia’s own Quantum-2 InfiniBand and Spectrum Ethernet.

Dell is also enhancing the AI features of the PowerEdge server lineup its PowerScale systems use for computing tasks. The upgrade is rolling out for the PowerEdge XE9680, an AI-optimized machine that debuted last year. It’s receiving support for the Instinct MI300X machine learning accelerator that Advanced Micro Devices Inc. launched on Wednesday.

AMD claims the MI300X can outperform Nvidia’s H100 HGX graphics card by 30%. The new accelerator comprises a dozen chiplets: Eight are GPUs based on a five-nanometer process, while the other four contain six-nanometer circuits that manage data input and output operations. There are also eight HBM3 memory modules for storing the information the MI300X processes.

PowerScale is one of two storage product lineups for which Dell introduced enhancements today. The other is the APEX Storage portfolio, which is receiving new tools designed to ease the management of file-based data. 

Dell is rolling out a new service called APEX File Storage for Microsoft Azure that companies can use to manage data they keep in Microsoft Corp.’s public cloud. Databricks Inc. customers, meanwhile, are receiving a set of integrations that will become available in AWS and Azure. The integrations will make it easier to train AI models on data a company manages using APEX file storage software. 

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU

Dell Technologies updates storage products for AI

Dell claims up to a 200% performance improvement for PowerScale systems.

Dell's latest move in the generative AI market is the introduction of the second-generation of its all-flash PowerScale storage array with smart scaleout capabilities, OneFS enhancements, and validation of Nvidia SuperPOD integration.

Dell's goal is to make the PowerScale an optimized storage solution for both general AI and generative AI workloads. It achieves this through new hardware and PowerScale OneFS software enhancements, so companies can prepare and infer AI models more quickly. It's promising a doubling of performance increase over the prior generation in both streaming reads and writes.

"The vast majority [of Dell customers] are planning to shift to an AI-first operating model, and in order to stay competitive in their in their separate spaces, how they’re going to do that is using their own data as the differentiator for their AI operations," said Martin Glynn, senior director of product management for Dell unstructured data storage solutions, on a conference call with journalists.

Glen said the latest projections for data growth are 175 zettabytes by the year 2025 and the majority of that is unstructured data in the form of images, videos, documents, and audio sensor data, which Dell kept in mind in designing PowerScale.

PowerScale will have a new scaleout architecture to improve single compute node performance and also to get around siloing of data, said Greg Findlen, senior vice president of product management for Dell's data management solutions.

"A lot of historical or legacy designs have really had coupled compute and storage from a scaling perspective and that has created a lot of limitations and burdens on our customers. [Proper storage] has got to make sure that it has a unified data stack. There’s a huge amount of unstructured data that’s sitting out there, and customers may need to be able to process that," he said.

Dell's strategy is to bring AI to wherever data resides, rather than the other way around. It notes that nearly 87% of companies are embracing multicloud strategies. Dell says it is giving customers the freedom to process data wherever it makes the most sense for them--on-premises, at the edge, or in public clouds.

While PowerScale is primarily an on-premises solution, Dell APEX File Storage in AWS and Azure allows customers to take advantage of cloud-native AI and GenAI workflows on data in public clouds as well as on-premises, with improved data access and movement.

Finally, through a collaboration with Nvidia, customers will have a validated combination of PowerScale with Nvidia's DGX systems, providing high-speed storage connected straight to Nvidia's AI hardware. Dell's solution is expected to be the first Ethernet storage solution validated on NVIDIA DGX SuperPOD. Dell PowerScale OneFS software enhancements will be globally available this month.

The new PowerScale systems will be available in the first half of 2024.

Join ARN

Copyright © 2024 IDG Communications, Inc. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of IDG Communications, Inc. is prohibited.

Foundry Sites: Computerworld Australia | CIO Australia | CMO Australia | CSO Online

Links: Privacy Policy [Updated 22 Feb 2023] | Copyright Notice |  | | Reprints | Advertising