- IBM Triples System Capacity to Support More Heavy Data Requirements of AI and Supercomputing
- The new all-flash chassis enables larger caches designed for dense multi-tenant cluster workloads.
- The advanced hardware is designed for operators scaling parallel processing pipelines for large data sets.
IBM has expanded the Storage Scale System 6000 to support full rack capacity up to 47 PB with the release of new All-Flash expansion enclosures equipped with 122 TB QLC flash drives.
This update represents a triple jump over previous limits and is designed for environments processing large data volumes.
The system is intended for organizations that work with supercomputing tasks, large artificial intelligence pipelines and cloud computing delivery of services.
Hardware designed for higher throughput
The company says the new design can handle workloads that rely heavily on consistent throughput and high availability.
It also said the larger platform makes it easier to scale for operators that support large clusters.
All-Flash Expansion Enclosure provides support for larger caches, allowing multi-tenancy at multiple levels within the cluster.
IBM says operators can run multiple data-intensive workloads without creating file system bottlenecks.
The housing can accommodate up to four Nvidia BlueField-3 DPU and twenty-six dual-port QLC flash drives in a 2U unit, enabling the system to meet the demands of AI training, simulation workloads, and extensive parallel processing.
Support for Nvidia Spectrum-X Ethernet switches is also included, reducing checkpointing time in model training processes.
IBM positions these hardware channels as essential in environments that require rapid data movement to support active GPU fleets and complex scheduling.
IBM has updated its Storage Scale System software to meet the growth in overall storage capacity.
Version 7.0.0 added support for higher capacity modules and enabled wider erase coding with a 16+2 configuration, designed to improve efficiency.
Write performance has also been increased to match the increase in throughput and IOPS, with earlier data for a four-rack configuration indicating system capacity of around 2.2 PB, up to 13 million IOPS, and read speeds of up to 330 GB per second.
The 2025 update raises the IOPS ceiling to 28 million and increases read throughput to 340 GB per second.
These changes are designed to ensure that expanded hardware does not introduce further delays as workloads scale.
The enclosure acts as a high-density option for operators who rely on solid state drive layer as the main storage base, while continuing to use cloud storage for distribution outside the main data center.
IBM says the increased capacity allows its global caching layer to store larger active data sets closer to GPUs, removing isolated islands of data and ensuring pipeline stability.
The architecture is built to serve clusters that require predictable movement of information between nodes, especially in situations where CPU load increases during heavy computing windows.
The company's announcement presents the update as a three-tier improvement that combines higher density, better data processing and broader workload support.
However, the long-term impact will depend on how consistently the system operates at full capacity after large-scale deployment.
By using HPCWire
Follow TechRadar on Google News. And add us as your preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the “Subscribe” button!
And of course you can also Follow TechRadar on TikTok for news, reviews, unboxing videos and get regular updates from us on whatsapp too much.
