AI workloads are generating massive amounts of data, with 90% being unstructured. RDMA for S3-compatible storage is being used to accelerate storage protocols for AI data. This solution, incorporating NVIDIA networking, offers faster object storage with lower costs and reduced latencies compared to TCP.

NVIDIA has developed RDMA client and server libraries to accelerate object storage, enabling faster data transfers for AI workloads. Several key object storage partners, including Cloudian, Dell Technologies, and HPE, are incorporating RDMA for S3-compatible libraries into their high-performance object storage products. Standardization efforts are underway to bring scalability and performance to existing S3-based applications.

Cloudian, Dell Technologies, and HPE are integrating RDMA for S3-compatible storage acceleration into their object storage products to meet the demands of AI workloads. NVIDIA’s RDMA for S3-compatible storage libraries will be available to select partners and are expected to be generally available in January via the NVIDIA CUDA Toolkit.

Cloudian, Dell Technologies, and HPE are collaborating with NVIDIA to integrate RDMA for S3-compatible storage into their high-performance object storage products, delivering faster, more efficient storage solutions for AI workloads. The libraries are expected to be available through the NVIDIA CUDA Toolkit in January.

Read more at NVIDIA: Unlocking Accelerated AI Storage Performance With RDMA for S3-Compatible Storage