The Storage Networking Industry Association (SNIA) has launched an open-standards project aimed at optimizing data services related to AI workloads.

The initiative, dubbed Storage.AI, plans to build an open ecosystem designed around making data services for AI workloads more efficient, targeting gaps in areas like processing and accessing data.

The non-profit’s members, including AMD, Cisco, and Pure Storage, have signed on to support the project. Industry ecosystem partners like the Ultra Ethernet Consortium and Open Compute Project are also involved, with technical work already underway.

"The unprecedented demands of AI require a holistic view of the data pipeline, from storage and memory to networking and processing,” Dr. J Metz, SNIA’s chair, explained.

“No single company can solve these challenges alone. SNIA’s Storage.AI provides the essential, vendor-neutral framework for the industry to coordinate a wide range of data services, building the efficient, non-proprietary solutions needed to accelerate AI for everyone."

Among the features SNIA members hope the project will support include unified software interfaces for better accessing nonvolatile memory express (NVMe) and remote direct memory access (RDMA) memory, as well as hybrid file and object storage backends capable of supporting scalability along the lines of Simple Storage Service (S3).

The group also plans to work on ways to bypass CPUs for data movement between GPU memory and RDMA/NVMe, which the nonprofit suggests would enable near-zero-latency input/output (I/O) to AI accelerators.

“The enterprise technologies powering today’s most demanding AI workloads are only as effective as the data they can access,” said Robert Hormuth, corporate VP of AMD’s data center solutions group. “We’re excited to collaborate with other leaders to give the industry a neutral, open path to fix data bottlenecks and unlock true AI performance.”

“As AI continues to expand across the enterprise, integrating robust storage solutions with AI workloads has become increasingly important,” added Jeremy Foster, SVP and GM of Cisco Compute.

“In today’s AI-driven landscape, data sets our customers apart, and the performance and reliability of that data are essential to realizing the full benefits of AI.”