Storage Built for Limitless Scale
No I/O bottlenecks. No storage compromises. ADS designs and integrates parallel storage solutions powered by IBM Storage Scale that feed your compute at full speed, from a single cluster to a global data platform.
Powered by IBM Storage Scale (GPFS) — the parallel file system trusted by the world's fastest supercomputers
Flagship Storage Appliance
Technical Specifications
Form Factor
Storage Controllers
System Memory
Drive Interface
Drive Technology
Raw Capacity
Sequential Bandwidth
Random IOPS
Network Connectivity
GPU Integration
File System
Protocols
Data Reduction
Ransomware Detection
Expansion
Management
4U (2 × 2U canisters, active-active)
Dual 48-core AMD EPYC "Genoa" — active-active HA
768 GB – 1.5 TB DDR5 RAM
PCIe Gen 5 x16 — 8 slots per canister (16 total)
IBM FlashCore Modules (FCM4) — NVMe, 19.2 TB or 38.4 TB
Up to 1.44 PB usable per node (TLC); 42+ PB/rack (QLC)
256 GB/s read per system (4U)
Up to 7 million IOPS per system
HDR/NDR InfiniBand or 100/200/400 GbE
NVIDIA GPUDirect Storage (GDS)
IBM Storage Scale (GPFS) — pre-installed and validated
POSIX / GPFS, NFS v4, SMB, S3, HDFS
Inline compression up to 2:1 (FCM4)
FCM4 ML-based anomaly detection
One or more external storage enclosures
IBM Storage Scale GUI, REST API, CLI
IBM Storage Scale
System 6000
The World's Fastest Parallel Flash Storage
The IBM Storage Scale System 6000 is a cloud-scale, high-performance parallel flash storage appliance purpose-built for AI training, HPC simulation, and advanced analytics. It delivers 2.5× the throughput of the nearest competitor in a dense 4U footprint — without the bottlenecks that throttle GPU workloads.
Powered by IBM Storage Scale software (the evolution of GPFS), the SSS 6000 serves as the high-speed data backbone connecting your GPU clusters to your entire data ecosystem — on-premises, hybrid cloud, and at the edge — under a single global namespace.
ADS is an authorized IBM Storage partner with deep expertise in sizing, integrating, and supporting SSS 6000 deployments alongside NVIDIA GPU platforms and HPC cluster environments.
IBM Storage Scale
IBM Storage Scale, formerly known as the General Parallel File System (GPFS), is the world's most proven high-performance parallel file system. It has powered the most demanding HPC and AI environments on the planet for over 25 years, from Fortune 500 data centers to national supercomputing facilities.
Unlike traditional NAS, Storage Scale distributes data and metadata across all nodes simultaneously, allowing thousands of compute clients to read and write in parallel at full aggregate bandwidth — with no metadata bottleneck, no single point of failure, and no practical limit to scale.
Trusted by the world's fastest computers. IBM Storage Scale (GPFS) is the file system for Summit at Oak Ridge National Laboratory, a 200 petaflops system with 9,000+ POWER9 processors and 27,000 NVIDIA Volta GPUs. It scales to exabyte capacity, billions of files, and thousands of concurrent compute nodes.
True Parallel I/O
All nodes read and write simultaneously across all drives. No metadata bottleneck. Throughput scales linearly with every added node and drive.
Exabyte Scalability
Start with tens of terabytes. Scale to exabytes without disruption or data migration. Over 1 billion files per filesystem, 8 EB namespace limit.
Enterprise Resilience
Synchronous mirroring, RAID-based erasure coding, journaling (write-ahead logging), and rapid non-disruptive recovery ensure data integrity at scale.
Global Unified Namespace
We start with your workload, not a product catalog. Every recommendation is based on your unique computational requirements.
Policy-Based Data Management
Automated ILM policies tier data across NVMe flash, SAS, SATA, tape, and cloud based on age, access frequency, file type, or custom rules.
Multi-Protocol Access
Simultaneous POSIX/GPFS, NFS v4, SMB, S3 object, and HDFS access, one data set, any application, any client OS, zero data copies.
Software Foundation
GPU-Accelerated I/O
NVIDIA GPUDirect Storage eliminates the CPU from the I/O path, creating a direct DMA path between Storage Scale and GPU memory for maximum throughput.
Hybrid Cloud & Edge
Extend your on-premises Storage Scale cluster to AWS, Azure, or IBM Cloud with transparent caching, tiering, and data mobility, no application changes.
Intelligent Data Tiering
IIBM Storage Scale's policy engine automatically places data on the right tier at the right time — maximizing performance for active workloads while minimizing cost for cold data, all under a single namespace.
NVMe Flash (IBM SSS 6000 / 3500) — Active training datasets, model checkpoints, HPC scratch space, and I/O-intensive workloads requiring maximum bandwidth and sub-millisecond latency. Fed directly to GPUs via GPUDirect Storage.
SAS / SATA SSD or High-Performance HDD (IBM SSS 5000) — Recently accessed datasets, intermediate results, staging for AI training queues. Automatically promoted to NVMe when access frequency increases.
High-Density Object Storage / Cloud — Infrequently accessed reference data, raw sequencing reads, historical simulation outputs. Available via S3 or NFS, with transparent on-demand recall into the hot tier.
Tape Library (IBM TS Series) or Cloud Archive — Long-term retention for regulatory compliance, raw data preservation, and disaster recovery. Storage Scale policy engine automates migration with full namespace visibility.
Workload Expertise
Optimized for Your Workload
Life Sciences & Genomics
Accelerate whole genome sequencing pipelines, variant calling, and bioinformatics workflows that generate and consume petabytes of FASTQ, BAM, and VCF files requiring high-throughput parallel I/O.
ADS has hands-on experience deploying IBM Storage Scale solutions across every major HPC and AI vertical. We know your workflows — and we know how to size storage to match them precisely.
AI Training & Inference
Feed multi-GPU clusters without starvation. IBM SSS 6000 with GPUDirect Storage delivers the sustained GB/s needed to keep NVIDIA GPUs at 100% utilization across multi-node inference and training runs.
Deliver high-performance computing to national laboratories, defense agencies, and research institutions with systems meeting compliance and security requirements.
Government & National Labs
University Research Computing
Power university research clusters with cost-effective, scalable CPU and GPU infrastructure backed by expert design and ongoing support.
HPC & Scientific Simulation
Parallel scratch storage for MPI applications. Storage Scale's distributed I/O architecture scales linearly with job size, whether you're running 100 or 10,000 compute processes.
Unify structured and unstructured data under a single namespace accessible via NFS, S3, and HDFS simultaneously.
Data Analytics & Data Lakes
Storage Expertise You Can Rely On
Deploying IBM Storage Scale correctly requires deep expertise. ADS brings decades of HPC storage experience to every engagement, from initial sizing through validation and ongoing support.
Workload-Driven Sizing
We analyze your actual I/O patterns, file sizes, access concurrency, protocol mix, and growth trajectory before recommending a single component. No generic configurations.
Validated Before Delivery
Every storage system is fully tested and benchmarked in our integration facility before it ships. You receive documented performance results against your actual workload profiles.
Workload-Driven Sizing
We analyze your actual I/O patterns, file sizes, access concurrency, protocol mix, and growth trajectory — before recommending a single component. No generic configurations.
IBM-Authorized Partnership
ADS is an authorized IBM Storage partner with certified expertise in Storage Scale. We have direct access to IBM technical resources and escalation paths for rapid issue resolution.
Full-Stack Integration
Storage doesn't exist in isolation. ADS designs compute, networking, and storage together, ensuring your InfiniBand fabric, GPU nodes, and Storage Scale cluster are balanced end-to-end.
Single Point of Support
One call covers the entire solution: storage hardware, Storage Scale software, networking fabric, and compute nodes. ADS handles all vendor coordination on your behalf.