IBM’s massive storage upgrade delivers a shock jump in raw density for teams pushing giant AI and supercomputing workloads
- IBM triples system capacity to support heavier AI and supercomputing data demands
- New flash enclosure enables larger caches designed for dense multitenant cluster workloads
- Expanded hardware targets operators scaling parallel processing pipelines across massive datasets
IBM has expanded the Storage Scale System 6000 to support a full rack capacity of up to 47PB, following the introduction of new All-Flash Expansion Enclosures equipped with 122TB QLC flash drives.
This update represents a threefold jump from earlier limits and is aimed at environments that handle high-volume data operations.
The system is positioned for organisations that work with supercomputing tasks, large AI pipelines, and cloud computing service delivery.
Hardware built for heavier throughput
The company claims the new design can sustain workloads that rely heavily on steady throughput and high availability.
It also states that the larger platform simplifies scaling for operators that maintain large clusters.
The All-Flash Expansion Enclosure brings support for bigger caches that enable multitenancy at multiple levels within a cluster.
IBM states operators can run several data-intensive workloads without creating bottlenecks across the file system.
The enclosure can house up to four Nvidia BlueField-3 DPUs and twenty-six dual-port QLC flash drives within a 2U unit, which allows the system to meet requirements linked to AI training, simulation workloads, and wide parallel processing.
Support for Nvidia’s Spectrum-X Ethernet switches is also included, allowing checkpoint times in model training processes to be shortened.
IBM positions these hardware links as essential in environments where fast data movement is needed to sustain active GPU fleets and complex scheduling.
IBM has updated its Storage Scale System software to align with the rise in total storage.
The 7.0.0 release adds support for the higher capacity modules and includes broader erasure coding with a 16+2 configuration that is intended to improve efficiency.
Write performance has also been increased to match the enhancements in throughput and IOPS, with earlier figures from the four-rack configuration placed the system at around 2.2PB of capacity, up to 13 million IOPS, and read speeds of up to 330GB per second.
The 2025 update lifts the IOPS ceiling to 28 million and raises read throughput to 340GB per second.
These adjustments aim to ensure that the expanded hardware does not introduce new delays when workloads scale.
The enclosure acts as a high-density option for operators that rely on an SSD layer as their primary storage base while continuing to use cloud storage for distribution beyond the core data centre.
IBM states that the increased volume allows its global-caching layer to keep larger active datasets closer to GPUs, removing separate data islands and keeping pipelines steady.
The architecture is built to serve clusters that need predictable movement of information between nodes, especially in situations where CPU utilisation rises during heavy compute windows.
The company’s messaging frames the update as a triple-tier improvement that combines higher density, better data handling, and wider workload support.
That said, the long-term impact will depend on how consistently the system performs at full capacity once deployed at scale.
Via HPCWire
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
