Menu
Data Management at Petabyte Scale for AI, HPC and Demanding File Workloads
IBM Storage Scale System is IBM's appliance-based deployment of IBM Storage Scale, combining scale-out file and object software with integrated NVMe hardware. It's built for organisations that need a high-performance, GPU-optimised global data platform — without building and tuning the software layer themselves.
IBM Storage Scale System is the hardware implementation of IBM Storage Scale — the software platform is available on its own too, but most organisations building an AI or HPC data platform want the pre-integrated appliance route rather than assembling and tuning the software layer themselves. The current family spans a compact, cost-efficient entry system through to a flagship platform built for national-lab and hyperscale performance.
Compact, cost-efficient entry system for AI pilots, HPC workloads, analytics and hybrid performance — with the same Storage Scale software foundation as the flagship 6000.
Enterprise-class platform for extreme AI, HPC and analytics workloads at national lab or hyperscale levels. NVIDIA-Certified, with up to 330 GB/s throughput per 4U system.
It's worth being clear on the distinction, since IBM uses both names. IBM Storage Scale is the software — a clustered, massively parallel file system with a global namespace, automated tiering and policy-driven data placement, available as a software-only deployment on your own hardware. IBM Storage Scale System is the appliance version — IBM's own hardware, pre-integrated with that same software, for organisations who want a faster route to production without building the platform themselves. If you're evaluating the software-only route, see our IBM Software page for more on how Storage Scale fits alongside the rest of the IBM portfolio.
Storage Scale System is built for organisations running data-intensive AI training, inference, HPC simulation or large-scale analytics — workloads where multiple compute nodes need fast, concurrent access to the same shared datasets. It's commonly deployed alongside GPU compute clusters, where storage throughput and latency directly determine how efficiently expensive GPU capacity gets used. If your workloads are more transactional or virtualisation-focused, IBM FlashSystem is likely the better starting point — the two product lines are often deployed together, with FlashSystem handling block workloads and Storage Scale System handling the unstructured, high-throughput file and object side.
If you're not sure which fits your workload, talk to Fortuna Data. As an IBM Silver Partner we can help you scope the right combination for your environment and budget.
