IBM has expanded its Storage Scale System 6000 to support up to 47PB of capacity in a single full rack, marking a threefold increase from its earlier limits. The upgrade follows the launch of new All-Flash Expansion Enclosures fitted with 122TB QLC flash drives, giving organisations a high-density option for handling large volumes of data. The development is aimed at operators running supercomputing workloads, large artificial intelligence pipelines, and cloud services that depend on high throughput and constant availability.
According to IBM, the new hardware design is built to deliver steady performance under heavy workloads while reducing the complexity of scaling large clusters. The expanded platform supports bigger caches that enable multitenancy across several layers of a cluster. IBM maintains that this helps operators run multiple data-intensive workloads simultaneously without causing bottlenecks across the file system.
The All-Flash Expansion Enclosure itself supports up to four Nvidia BlueField-3 DPUs and twenty-six dual-port QLC flash drives within a compact 2U frame. This combination is designed for tasks linked to AI training, simulation workloads, and wide parallel processing. Support for Nvidia’s Spectrum-X Ethernet switches is also included, which IBM says can shorten checkpoint times during model training. The company highlights these hardware connections as vital for environments that need fast data movement to keep GPU fleets active and to manage complex scheduling.
Software updates matched to rising capacity
To support the expanded hardware, IBM has updated its Storage Scale System software. The latest 7.0.0 release enables the system to work with higher-capacity modules and introduces broader erasure coding with a 16+2 configuration, which is intended to improve efficiency and durability. Write performance has increased in line with upgrades to throughput and IOPS, ensuring the software keeps pace with the hardware.
Earlier figures for the four-rack configuration cited capacity at around 2.2PB, delivering up to 13 million IOPS and read speeds of up to 330GB per second. The 2025 update significantly raises the performance ceiling, lifting IOPS to 28 million and increasing read throughput to 340GB per second. IBM notes that these improvements are meant to prevent delays as operators scale their systems and run heavier workloads on the expanded platform.
The updated enclosure supports operators that rely on an SSD layer as their primary storage tier, while still using cloud platforms for distribution beyond the central data centre. IBM states that the greater storage volume allows its global-caching layer to hold larger active datasets closer to GPUs. This removes isolated data silos and helps maintain a steady data pipeline for demanding tasks.
Designed for predictable data movement
IBM says the system architecture is created for clusters that need predictable, controlled movement of information between nodes, especially when CPU usage rises during periods of intensive computing. By combining higher density, improved data handling, and wider workload support, the company describes the update as a triple-tier improvement to its existing platform.
However, the long-term impact will depend on how reliably the system performs once deployed at full capacity. As organisations undertake larger data operations and more complex scheduling, real-world performance will determine how well the Storage Scale System 6000 meets future demands.
