Description

DDN EXAScaler is a flagship family of high-performance parallel file system appliances engineered to meet the extreme demands of artificial intelligence (AI), high-performance computing (HPC), and data analytics. Built on the industry-leading Lustre-based EXA6 architecture, EXAScaler delivers unmatched throughput, ultra-low latency, and massive concurrency for GPU-accelerated environments.

Designed for seamless integration with NVIDIA DGX BasePOD and SuperPOD deployments, EXAScaler supports GPU Direct Storage (GDS), Hot Nodes caching, and end-to-end NVMe data paths—enabling direct communication between NVMe SSDs and GPUs without CPU bottlenecks. This architecture dramatically reduces IO latency and accelerates deep learning, generative AI, and simulation workloads.

EXAScaler appliances are available in multiple form factors and configurations, allowing organizations to scale from departmental AI clusters to enterprise-grade, multi-tenant GPU platforms. With support for InfiniBand HDR/NDR, 200/100GbE, and advanced data orchestration tools like Stratagem and Hot Pools, EXAScaler ensures optimal performance, data integrity, and operational efficiency across hybrid and cloud environments.

EXAScaler Product Family – Model Comparison

Model	Form Factor	Target Use Case	Key Features
EXAScaler EXA6	2U / 4U Appliance	DGX BasePOD, SuperPOD, Enterprise AI	Parallel Lustre, GDS, Hot Nodes, 115 GB/s read, 3M+ IOPS
EXAScaler ES200X2	2U Appliance	Departmental AI, Edge Inference	NVMe-oF RoCE, FlashLink, 0.05ms latency, scalable IOPS
EXAScaler ES400X2	4U Appliance	Mid-range AI Training	SmartMatrix full-mesh, 32 controllers, 100GbE RDMA
EXAScaler SFA400X2	4U Appliance	Mixed AI/HPC Workloads	NVMe block storage, hybrid expansion, high SSD density
EXAScaler GRIDScaler	Multi-node	Enterprise-scale AI, Multi-tenant clusters	Parallel Lustre (EXA6), GDS, 8× HDR IB, 115 GB/s read, validated with DGX H100

Key Architectural Highlights

Parallel Lustre File System (EXA6): Fully optimized for AI/HPC workloads with scalable metadata and data paths.
GPU Direct Storage (GDS): Enables direct NVMe-to-GPU data transfer, reducing latency and CPU overhead.
Hot Nodes Caching: Automatically caches frequently accessed data on GPU-local NVMe, accelerating IO.
Stratagem Policy Engine: Advanced data placement and residency control for hybrid and multi-cloud environments.
Hot Pools Tiering: Intelligent movement of data between flash and disk tiers for optimal performance and cost-efficiency.
T10DIF Data Integrity: Ensures end-to-end protection from application to disk.
InfiniBand HDR/NDR & 200/100GbE: High-bandwidth, low-latency connectivity for AI pipelines and GPU clusters.
Management Framework (EMF): Modern CLI/API orchestration, observability, and automated upgrades across clusters.

Strategic Positioning for TIADA

DDN EXAScaler represents the pinnacle of AI-optimized storage infrastructure. For TIADA’s enterprise clients, it offers:

Validated integration with NVIDIA DGX H100 systems
Scalable performance for multi-tenant AI platforms
Turnkey deployment with full orchestration and observability
Competitive edge over traditional NAS and block storage solutions like Huawei Dorado or Pure Storage

Main Models of DDN EXAScaler EXA6

1. DDN ES/AI200X2

Compact, entry-level appliance
Ideal for Edge AI, departmental inference, Dev/Test environments
Scale-out architecture with Lustre EXA6 support
GPU hot-node caching support

2. DDN ES/AI400X2

Mid-range model designed for DGX BasePOD deployments
Scales linearly with additional appliances
Performance: 90+ GB/s Read, 65+ GB/s Write, ~3M IOPS
Fully validated with NVIDIA DGX H100 systems
Supports GPU Direct Storage (GDS)

3. DDN SFA400X2

Pure NVMe block storage for parallel file systems
Scale-out building block for large AI pipelines
Suitable for hybrid expansion and high-throughput AI workloads
Delivers strong throughput and IOPS scalability

4. DDN GRIDScaler (IBM Spectrum Scale-based)

Enterprise-grade Parallel File/NAS appliance
Built on IBM Spectrum Scale (GPFS)
Designed for multi-tenant AI/HPC platforms
Delivers high throughput with 0.61ms SPEC SFS2014 response time
GPU integration paths for advanced AI workloads

Comparison Table: DDN EXAScaler EXA6 Models

Model	Form Factor / Interfaces	Performance	AI/HPC Use Case	Special Features
ES/AI200X2	2U appliance, 4× HDR/HDR100 IB or 4× 200/100GbE	High IOPS, compact	Small GPU clusters, Dev/Test, Edge AI	Lustre EXA6, Hot Nodes GPU caching
ES/AI400X2	2U appliance, 8× HDR/HDR100 IB or 8× 200/100GbE	90+ GB/s Read, 65+ GB/s Write, ~3M IOPS	DGX BasePOD, mid-size AI training clusters	NVIDIA DGX H100 validated, GDS support
SFA400X2	Block storage, 8× HDR/HDR100 IB or 8× 200/100GbE	High throughput; scalable	Large-scale AI pipelines, hybrid expansion	Pure NVMe block storage, filesystem backend
GRIDScaler	Parallel NAS/File appliance with IBM Spectrum Scale	SPEC SFS2014: 0.61ms response, high throughput	Enterprise-scale, multi-tenant AI/HPC	GPU integration, end-to-end parallel paths

ddn

DDN EXAScaler – Scalable Parallel File System Appliances for AI, HPC, and Data-Intensive Workloads

Description

EXAScaler Product Family – Model Comparison

Key Architectural Highlights

Strategic Positioning for TIADA

Main Models of DDN EXAScaler EXA6

1. DDN ES/AI200X2

2. DDN ES/AI400X2

3. DDN SFA400X2

4. DDN GRIDScaler (IBM Spectrum Scale-based)

Comparison Table: DDN EXAScaler EXA6 Models

Brand

ddn

Reviews

Shipping & Payment

Additional information

Related products

Huawei OceanStor Dorado 6000 V6 – All-Flash NVMe Storage System for Mission-Critical Enterprise Applications

NVIDIA H100 80GB PCIe Tensor Core GPU

NVIDIA RTX A6000

NVIDIA L40 GPU

H3C S9827 Series High-Density Intelligent Data Center Switches

H3C S9855 Series High-Density RoCEv2 Ethernet Switch for AI & HPC Data Centers