IBM introduced key enhancements to its IBM Storage Scale System 6000, designed to power data-intensive AI workloads by eliminating data silos and accelerating access across edge, core, and cloud environments.
IBM emphasizes that data unification is fundamental to successfully operating an “AI factory.” The Storage Scale System 6000 now leverages a global namespace and Active File Management (AFM) to unify data, remove storage silos, and deliver rapid access for AI pipelines.
With its latest release, IBM has tripled the maximum capacity of the Storage Scale System 6000 to 47 petabytes per rack, by introducing support for industry-standard QLC flash in 30 TB, 60 TB, and 122 TB SSD configurations.
Software updates (Storage Scale System 7.0.0) introduce:
Multi-flash tiering, enabling both NVMe TLC and QLC flash drives in a single system
A data acceleration tier built on NVMe-oF, delivering up to 340 GB/s throughput and up to 28 million IOPS
Enhanced data protection via a broader erasure coding framework (16+2/3P) for better disk efficiency and write performance
Optimized training and AI model checkpointing via NVIDIA Spectrum-X Ethernet networking, reducing checkpointing time for foundation models
In addition, IBM is launching a new All-Flash Expansion Enclosure tailored for high-performance AI training, inference, HPC, and large-scale data operations. When paired with the 122 TB QLC NVMe SSDs, the system delivers over 47 PB of dense flash capacity in a 42U configuration.
Also Read: Microsoft and NVIDIA Partner on Real-Time Cyber Immunity
This expansion enclosure integrates with NVIDIA BlueField-3 DPUs, offering:
Up to 400 Gb/s of network throughput with NVMe-oF offload
Over 3 PB of raw flash storage in a compact 2U design.
Supports four BlueField-3 DPUs and 26 dual-port QLC flash drives. This enables high-performance, multi-tenant, and secure deployments.
IBM claims these upgrades help service providers, AI companies, and HPC environments manage multi-tenant workloads better. They also ensure data isolation, performance, and security.
The IBM Storage Scale System 6000 seeks to remove bottlenecks. It does this by tripling capacity and using a high-density cache close to GPU infrastructure. It speeds up AI pipelines. Organizations can also collaborate on data from anywhere, regardless of format or location.