Expanding AI Storage Access with NVIDIA cuObject and the NVIDIA SCADA Server SDK

NVIDIA announced the general availability of its cuObject client and server libraries, which provide standardized APIs and an RDMA wire protocol for accelerated object storage access. According to NVIDIA, the libraries let AI accelerators reach file and object storage over RDMA without routing data through the server CPU, which the company says enables higher throughput, lower latency, and reduced CPU utilization for writes and reads.

NVIDIA is also expanding the xio-sig consortium to include cuObject alongside cuFile, in partnership with Google Cloud and Microsoft. Google Cloud, already a maintainer for cuFile in xio-sig, is evaluating expanding its participation for cuObject. Microsoft looks forward to joining the xio-sig Board to improve interoperability in storage I/O. NVIDIA said the repository structure is set up for both cuFile and cuObject, and that headers, the cuObject wire protocol, and libxFile and xFilekernel implementation code will be shared once the production-ready stack passes conformance tests. Governance documents are under review by pending board members.

The new Scaled Accelerated Data Access (SCADA) Server SDK enables storage providers to build SCADA servers that receive requests from GPU-based SCADA clients, fulfill them using local or remote storage, and deliver results over RDMA. The effort also includes a Storage Lender Service and a SCADA command-line utility. IBM Storage has demonstrated a prototype in which a SCADA client sends requests to its initial version of the Storage Scale SCADA server, built on the new SDK. NVIDIA said this shows how storage vendors can collaborate on an ecosystem for GPU-initiated storage access, with potential applications including semantic search, recommender systems, and fraud detection.

Through the NVIDIA Storage-Next initiative, NVIDIA said it is leading a group of over 40 vendors and customers, including NAND vendors, controller vendors, storage providers, hyperscalers, and application developers, to define open industry standards for GPU-driven, fine-grained storage access, with SCADA as the supporting software infrastructure. NVIDIA pointed developers to the xio-sig repository for updates, cuObject documentation, and cuObject Server 2.0.0 for evaluation.