Key Highlights

  • NVIDIA’s RDMA for S3-compatible storage accelerates AI data transfer by up to 90%
  • Scalable storage solutions for AI workloads, reducing costs and increasing efficiency
  • Partners like Cloudian, Dell Technologies, and HPE are adopting the new technology

The increasing demand for artificial intelligence (AI) and machine learning (ML) applications has led to an explosion of data generation, with enterprises projected to produce nearly 400 zettabytes of data annually by 2028. This massive scale, combined with the need for data portability between on-premises infrastructure and the cloud, has pushed the AI industry to evaluate new storage options. NVIDIA’s introduction of RDMA for S3-compatible storage is a significant development in this space, enabling faster and more efficient object storage for AI workloads.

The Need for Scalable Storage

The traditional TCP network transport protocol is no longer sufficient for the high-performance requirements of AI applications. RDMA for S3-compatible storage addresses this issue by using remote direct memory access (RDMA) to accelerate S3-API-based storage protocols. This results in higher throughput per terabyte of storage, lower latency, and reduced CPU utilization. As Jon Toor, chief marketing officer at Cloudian, notes, “Object storage is the future of scalable data management for AI.” The benefits of RDMA for S3-compatible storage include:

  • Lower cost per terabyte
  • Higher throughput per watt
  • Significantly lower latencies compared to TCP
  • Improved workload portability between on-premises and cloud environments

Industry Adoption and Standardization

NVIDIA is working with partners to standardize RDMA for S3-compatible storage, with several key object storage partners already adopting the new technology. Cloudian, Dell Technologies, and HPE are integrating RDMA for S3-compatible libraries into their high-performance object storage products. As Rajesh Rajaraman, chief technology officer and vice president of Dell Technologies Storage, Data and Cyber Resilience, comments, “AI workloads demand storage performance at scale with thousands of GPUs reading or writing data concurrently.” The widespread adoption of RDMA for S3-compatible storage is expected to drive innovation and growth in the AI industry.

Conclusion and Future Developments

The introduction of RDMA for S3-compatible storage marks a significant milestone in the development of scalable and efficient storage solutions for AI workloads. As the AI industry continues to evolve, the need for high-performance storage will only continue to grow. With NVIDIA’s RDMA for S3-compatible storage libraries now available to select partners, we can expect to see further advancements in this space. As Jim O’Dorisio, senior vice president and general manager of storage at HPE, notes, “NVIDIA’s innovations in RDMA for S3-compatible storage APIs and libraries are redefining how data moves at massive scale.”

Source: Official Link