Pure Storage announces FlashBlade//EXA data storage platform for HPC and AI computing

Pure Storage has announced they are going to launch the FlashBlade//EXA data storage platform during mid-2025 to serve the needs of AI and HPC.

Built for high concurrency and large-scale metadata operations, it is capable of delivering more than 10TBps read performance within a single namespace while traditional solutions still struggle to break that bottleneck.

Due to the nature of AI model training that involves diverse data types like text, images, and videos, Pure storage utilizes advanced metadata engine and Purity operating system to enable a massive parallel processing stack that is also scalable to ensure the best reliability, reducing idling times, and accelerate overall AI model training and inference times.

FlashBlade//EXA, built on a decade of Pure Storage innovation, will:

• Deliver industry-leading performance at scale, optimizing AI pipeline efficiency and minimizing training and inference delays.
• Provide multi-dimensional performance with massive parallel processing and scalable metadata IOPS, offering over 10 terabytes per second in a single namespace.
• Reduce complexity by eliminating metadata bottlenecks, ensuring high metadata availability and resiliency for large AI datasets without manual tuning.
• Support AI and HPC innovation with a configurable, disaggregated architecture leveraging industry-standard protocols and NVIDIA’s high-speed networking technology.

FlashBlade//EXA is poised to offer independent scaling for data and metadata, supports near-unlimited expansion with off-the-shelf third-party data nodes, and simplifies deployment, management, and scaling through standard protocols and networking.