Achievements in AI storage performance verified through MLPerfâ Storage V2 results by Western Digital

Achievements in AI storage performance verified through MLPerfâ Storage V2 results by Western Digital

Western Digital's OpenFlex Data24 4000 Series Offers High-Performance, Scalable AI Storage

Western Digital's OpenFlex Data24 4000 Series NVMe-oF Storage Platform is designed to meet the growing demands of modern AI infrastructure. This solution offers high-performance, scalable, low-latency shared storage, enabling independent scaling of storage and compute resources to meet complex AI workloads efficiently and cost-effectively.

Key features and benefits include:

1. NVMe over Fabrics (NVMe-oF) technology: The OpenFlex Data24 utilizes NVMe-oF technology, which extends high-speed NVMe flash storage over Ethernet, allowing for rapid data access with low latency crucial for AI processing.
2. Scalability and flexibility: The platform supports scaling of storage capacity and compute (GPU) resources independently, accommodating growing GPU demand in AI training and inference environments for flexible and efficient resource utilization.
3. Validated performance with MLPerf Storage V2 benchmarks: Western Digital's OpenFlex Data24 has demonstrated optimal real-world AI workload performance and efficiency across large simulated GPU deployments. For instance, it delivered outstanding performance across 186 simulated H100 GPUs and three client nodes, with an impressive GPU-to-drive ratio.
4. Simplified deployment and cost-effectiveness: The OpenFlex Data24 is designed to reduce complexity in AI infrastructure setups while providing a high GPU-to-drive ratio, leading to better cost-performance balance compared to traditional storage solutions.
5. Collaboration with PEAK:AIO AI Data Server: Western Digital collaborated with PEAK:AIO, a high-performance SDS provider, to enhance overall AI data throughput and server integration for smoother AI workload operations. With the PEAK:AIO AI Data Server, OpenFlex Data24 was able to saturate 52 simulated H100 GPUs from a single head server and single client node.

When deployed in the OpenFlex Data24 enclosure, KIOXIA CM7-V Series NVMe SSDs enable sustained, high-performance disaggregated data delivery to many GPU client-nodes. The validation submission utilized KIOXIA CM7-V Series NVMe SSDs, known for their outstanding performance in demanding AI workloads.

Roger Cummings, President and CEO at PEAK:AIO, stated that combining PEAK:AIO's software-defined AI Data Server with the scalability of Western Digital's OpenFlex Data24 and the performance density of KIOXIA's CM7-V Series SSDs delivers high-performance AI infrastructure that's faster to deploy, more efficient to operate, and easier to scale.

Kurt Chan, vice president and general manager, Western Digital Platforms Business, stated that the OpenFlex Data24 maximizes GPU utilization while minimizing footprint, complexity, and overall total cost of ownership.

In summary, the OpenFlex Data24 4000 Series is engineered to meet modern AI infrastructure needs with high throughput, scalability, low latency, and efficient resource utilization, making it a robust and validated solution for managing AI storage demands at scale.

Technology, such as the NVMe over Fabrics (NVMe-oF), plays a significant role in Western Digital's OpenFlex Data24 4000 Series, a solution for data-and-cloud computing that extends high-speed NVMe flash storage over Ethernet. Furthermore, the technology fosters collaboration with other providers, as demonstrated by Western Digital's partnership with PEAK:AIO, aiming to deliver high-performing AI infrastructure.

Read also: