IBM Storage Scale 6.0.1 Improves Storage for AI and HPC
Unstructured corporate data is spread across distributed file and object storage environments located in data centers, cloud storage, and edge sites. IBM Storage Scale breaks down these data silos and makes dispersed unstructured data AI‑ready, delivering high bandwidth and ultra‑low latency access to data‑hungry GPUs for AI at scale. It can be deployed as primary storage for large AI Factories or attached to existing primary storage to enable secure AI insights from private corporate data.
IBM Storage Scale evolves into the IBM Storage Scale AI Data Platform by integrating IBM Content Aware Storage (CAS). CAS is a revolutionary capability that transforms unstructured, dispersed data into AI‑ready data (vectorized on ingest) and securely provisions it to AI inferencing applications without the data leaving the organization. The accuracy of vectorized data increases the value of produced tokens.
The IBM Storage Scale AI Data Platform supports the full AI lifecycle – from data ingestion through model training and adaptation to inference.
AI-Accelerated Storage Architecture
By integrating with NVIDIA GPUs, DPUs, and AI software, the IBM Storage Scale AI Data Platform transforms enterprise storage into a high‑throughput engine for data ingestion, embedding, and retrieval – powering model training, adaptation, and inference at scale.
Seamless Integration with NVIDIA AI Infrastructure
NVIDIA has certified the IBM Storage Scale AI Data Platform for NVIDIA DGX BasePOD, NVIDIA DGX SuperPOD, and NVIDIA Cloud Platform reference architectures, ensuring enterprise‑grade performance and reliability across the entire AI workflow, including agentic AI workloads.
Near Real-Time, Secure Provisioning of Private Data
Built on IBM Storage Scale, IBM Content Aware Storage, and IBM Fusion Data Catalog, the platform automatically transforms unstructured data when it is created or updated on primary storage. It then securely provisions the data – without copying and without delay – to AI inferencing applications, ensuring governance and compliance
Fast Data Access for Thousands of Tenants
Designed for flexibility and rapid AI insights, the Storage Scale AI Data Platform supports high‑performance data access to GPU servers, including NVIDIA GPUDirect Storage and NVIDIA cuObject (S3 over RDMA), for up to 3,000 securely isolated tenants.
What’s New in Storage Scale 6.0.1
Storage Scale 6.0.1 introduces numerous enhancements to better support data‑intensive workloads, including AI, HPC, analytics, and data‑driven science and engineering.
Storage for AI & HPC
- Validated solution architectures for scalable, high‑volume inferencing, including integration with IBM Content Aware Storage
- Validated solution architectures for GPU‑accelerated databases
- Updated NVIDIA certifications
- NVIDIA Nsight integration
Multi-Tenancy
- Support for up to 3,000 tenants, each with dedicated encryption keys
- New GUI for tenant management and tenant provisioning workflows
- Updated multi-tenancy reference architecture
Data Mobility
- AFM Watch Folder support for NFS targets to improve efficiency of caching
- Improved AFM scalability and resilience for NFS targets
- cuObject (S3 over RDMA) support, achieving up to 90% of the GPFS POSIX performance
- S3 versioning
- S3 IAM user management
- SMB Multichannel
Serviceability
- Improved problem determination for AFM, protocols, snapshots, and networking
Deployment & Configuration
- Enhanced GUI capabilities
- S3 bucket configuration improvements
- Wizard for AFM caching (prefetch and eviction)
- Enhanced API‑driven control plane
References
Storage Scale 6.0.1 Documentation
IBM Content Aware Storage
GPU accelerated databases
NVIDIA Storage Solutions
Multi-tenancy
It is planned to publish blog posts with more technical details. This section of references will be updated as more blog posts get published.