Pure Storage, the pioneering IT company that offers the world’s most advanced data storage technology and services, today unveiled FlashBlade//EXA™, the industry’s highest-performing data storage platform designed to meet the demanding requirements of artificial intelligence and high-performance computing (HPC). While legacy data storage approaches have hindered AI’s potential, FlashBlade//EXA addresses the metadata bottleneck with a proven architecture based on FlashBlade, built for high concurrency and the massive amounts of metadata operations typical of large-scale AI and HPC workloads. In preliminary testing, FlashBlade//EXA is projected to deliver read performance of over 10 terabytes per second in a single namespace, setting a new standard as the industry’s highest-performing storage solution.
The FlashBlade//EXA architecture independently scales data and metadata; provides near-unlimited scaling, with commercially available off-the-shelf (OTS) data nodes that enable highly scalable multidimensional performance; and reduces complexity in deployment, management, and scaling by utilizing standard protocols and networks.
Driving a Paradigm Shift in Storage
The most powerful GPUs have accelerated the pace and scale at which foundational AI models can be trained. The explosive growth in the size and sophistication of models is driving a paradigm shift in storage requirements, where solutions must keep up with the computational intensity and volume, as well as the diverse data demands of AI and HPC.
Legacy storage systems were not designed to meet modern AI requirements. When applied to large-scale AI and HPC, they face critical limitations with parallel and simultaneous reads and writes, metadata performance, ultra-low latency, asynchronous checkpoints, and high, predictable transfer rates.
A modern storage solution must provide a massively parallel and disaggregated architecture to offer flexible scaling, ensuring that storage helps accelerate the pace of AI.
A Modern Storage Architecture for AI and HPC Workloads
Legacy high-performance storage architectures were optimized for traditional HPC environments, with more predictable and regular workloads, focusing on scaling raw performance. Current AI workloads are complex and multimodal (including text, images, and videos), processed simultaneously by tens of thousands of GPUs. This drastic shift demands advanced metadata optimization alongside massive performance scaling to effectively manage diverse data types and high concurrency.
FlashBlade//EXA is specifically designed to tackle the challenges of AI workloads with unmatched performance and metadata management. Its disaggregated and massively parallel architecture enables large-scale storage flexibility. Organizations can adapt to the evolution of multimodal models, optimize reliability, and eliminate downtime to accelerate training and inference of AI models while enhancing GPU utilization. Furthermore, the combination of Pure Storage’s metadata engine and the Purity operating system, alongside cost-efficient off-the-shelf data nodes, enables businesses to achieve an unprecedented price-performance ratio.
FlashBlade//EXA builds on a decade of innovation from Pure Storage to provide transformative storage for AI and HPC environments. FlashBlade//EXA will enable:
- Deliver industry-leading performance at scale, leveraging Pure Storage’s proven metadata capabilities to maximize AI channel efficiency and minimize delays during training and inference workflows.
- Enable multidimensional performance with massive parallel processing and scalable metadata IOPS to meet high-performance AI requirements. Provide over 10 terabytes per second in a single namespace, representing the best performance in the industry.
- Simplify management complexity by eliminating metadata bottlenecks
- Deliver high performance, availability, and resilience of metadata to manage massive AI datasets without manual tuning or additional configurations.
- Accelerate AI innovation with a highly configurable and disaggregated architecture
- Drive the changing landscape of AI and HPC with industry-standard protocols.
- Incorporate high-speed NVIDIA ConnectX NICs, Spectrum switches, LinkX cables, and accelerated communication libraries.