IBM Storage Scale System 6000
Storage Built for Limitless Scale
Parallel flash storage for AI, HPC, and data-intensive research
ADS designs IBM Storage Scale solutions to deliver sustained parallel I/O from a single high-performance system to a global data platform — with file, object, and GPU-adjacent access under one architecture.
IBM Storage Scale is the evolution of GPFS, a parallel file system with decades of deployment in demanding supercomputing environments.
Request a Consultation
Download Datasheet
Flagship Storage Appliance
IBM Storage Scale System 6000
A dense, active-active parallel flash appliance built for workloads that need high sustained bandwidth, high IOPS, and shared access from large compute clusters.
Storage is integrated with the compute and network fabric so throughput is designed end-to-end rather than evaluated in isolation.
ADS sizes the storage platform around real I/O patterns — file size, concurrency, protocol mix, growth, and GPU data movement.
Form Factor
4U system using two 2U active-active canisters
Storage Controllers
Dual AMD EPYC-based active-active controllers
System Memory
768 GB to 1.5 TB DDR5 RAM
Drive Technology
IBM FlashCore NVMe modules with high-capacity options
Capacity
Up to ~1.44 PB usable per system in TLC configurations; denser QLC rack options
Sequential Bandwidth
Up to 256 GB/s read per 4U system in current published configurations
Random IOPS
Up to 7 million IOPS per system
Connectivity
HDR/NDR InfiniBand or 100/200/400 GbE
GPU Integration
NVIDIA GPUDirect Storage
Protocols
POSIX/GPFS, NFS, SMB, S3, and HDFS access options
Data Services
Compression, policy management, monitoring, and anomaly detection
Management
IBM Storage Scale GUI, REST API, and CLI
Software Foundation
IBM Storage Scale
A parallel data platform that distributes I/O and metadata activity across the cluster, allowing many clients to access shared data concurrently without depending on a traditional single-controller NAS model.
True Parallel I/O
Clients read and write across multiple storage resources simultaneously, allowing aggregate throughput to scale with the system.
Scale Without a Forklift
Capacity and performance can expand over time while preserving a single namespace for users and applications.
Enterprise Resilience
Data protection, journaling, redundancy, and recovery features are designed for high-availability technical computing.
Global Unified Namespace
Data can be presented consistently across local, remote, and hybrid environments.
Policy-Based Data Management
Automated policies can place data on flash, disk, object, tape, or cloud tiers according to age, access, or business rules.
Multi-Protocol Access
The same data can be exposed through multiple file and object protocols without maintaining separate copies.
Intelligent Data Tiering
Put Each Data Set on the Right Tier
Storage Scale policies can keep active training and simulation data on the fastest media while moving colder content to more economical tiers under the same namespace.
HOT TIER
NVMe Flash
Active training data, checkpoints, scratch space, and latency-sensitive I/O can live on SSS 6000-class flash and feed GPUs through GPUDirect Storage.
WARM TIER
SSD / High-Performance HDD
Recently used data, staging areas, and intermediate results can remain online at a lower cost than all-flash capacity.
CAPACITY TIER
High-Density Object / Cloud
Large reference data sets, raw sequencing reads, and historical results can be moved to lower-cost capacity while remaining discoverable.
ARCHIVE
Tape or Cloud Archive
Long-term preservation and compliance data can be moved automatically while namespace visibility is retained.
Workload Expertise
Optimized for the Data Path
ADS has experience aligning Storage Scale with compute-heavy workflows where data access directly affects application throughput.
Life Sciences & Genomics
Parallel access for large sequencing pipelines, genome analysis, and data sets containing many FASTQ, BAM, and VCF files.
AI Training & Inference
Sustained throughput and GPUDirect Storage for multi-GPU training, checkpointing, and inference data.
Government & National Labs
High-performance shared storage for demanding scientific and research environments.
University Research Computing
Scalable storage for shared campus clusters serving diverse applications and research teams.
Scientific Simulation
Parallel scratch and output paths for MPI applications from hundreds to thousands of compute processes.
Analytics & Data Lakes
Unified file and object access for large unstructured data sets and analytics pipelines.
Eliminate the Storage Bottleneck
Ready to Size an IBM Storage Scale Platform?
ADS can model capacity, IOPS, bandwidth, protocol access, and growth so the storage architecture stays aligned with your compute environment.
Request a Consultation
Applied Data Systems
©2026 Applied Data Systems