NetApp plans to acquire PEAK:AIO, adding high-performance parallel file-system and metadata technology designed to scale storage alongside increasingly large AI GPU clusters.

The Sept. 25 agreement targets one of the less visible constraints inside an AI factory: keeping accelerators supplied with data as file counts, concurrent jobs and storage capacity grow.

Parallel metadata for trillions of files

NetApp says it plans to combine PEAK:AIO’s specialized metadata services and parallel namespace technology with ONTAP. The architecture is intended to support trillions of files and multi-exabyte deployments while preserving ONTAP’s data-management foundation.

PEAK:AIO separates metadata activity from the primary data path through parallel NFS. That allows metadata services and storage capacity to scale more independently, which can reduce a bottleneck that becomes increasingly important as AI workloads create enormous numbers of files and simultaneous I/O requests.

The embedded English-language StorageReview video features PEAK:AIO founder Mark Klarzynski discussing the architecture on a Dell PowerEdge R7725xd. StorageReview reports approximately 300 GB/s inside the server, 160 GB/s across the network and roughly 70 million IOPS in its demonstration.

Storage has to keep GPUs busy

Large GPU clusters can lose useful compute time when storage cannot deliver training data, checkpoints, model weights or inference context quickly enough. NetApp says PEAK:AIO’s technology is intended to reduce data-related GPU stalls while retaining standards-based parallel NFS access.

The technology originated through work involving Los Alamos National Laboratory and Carnegie Mellon University. PEAK:AIO has also been developing its Lattice open-source pNFS metadata architecture for large IT, AI and HPC environments.

NetApp extends ONTAP toward AI factories

The proposed acquisition would give existing ONTAP environments an evolution path toward a more distributed metadata layer instead of requiring every AI deployment to adopt an entirely separate storage architecture. The transaction remains subject to customary closing conditions and regulatory approval.

The announcement follows BitcoinVersus.tech coverage of Clearfield’s 2,304-fiber AI data-center panel, Marvell’s 2 nm optical networking and Giga Computing’s 700 kW GAIFA AI factory.

Sources: NetApp and StorageReview.

Twitter:

BitcoinVersus.Tech Editor’s Note & Disclaimer:
We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. BitcoinVersus.tech provides news, technical analysis and educational information and does not provide financial or investment advice. Verify technical specifications and operational requirements with the manufacturer before making purchasing, deployment or investment decisions. If you would like to support to help further secure the integrity of our research initiatives, please donate here: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb