PRODUCT

Organizations leading the future with endless challenges and changes, organizations delivering fast and convenient

AI/HPC Computing

 

LustreParallel file system

DDN Storage powers the world's most powerful HPC and AI compute resources.

Supports the world’s most data-intensive
AI workloads and largest ML/DL datasets

1

NVIDIA Eos Next-Generation DGX SuperPOD

4608 H100 GPU (DGX H100 x576)
expected to provide 18.4 Exaflops AI computing performance

2

NVIDIA SELENE Supercomputer Top#09

4500+ A100 GPUs and 13 PB DDN storage

3

NVIDIA CAMBRIDGE-NVIDIA CAMBRIDGE-1

400+ PETAFLOPS w/ DDN Storage

NVIDIA DGX POD™ has the biggest number of used case

NVIDIA selected DDN Storage as the most suitable storage for DGX POD configuration, and as the storage for NVIDIA's most powerful supercomputer.

  • Building a hyper-scale GPU cluster DGX SuperPOD
  • 6M SMS/day, message analysis
  • 120K, real-time cameras in smart cities
  • 640 PB+, autonomous vehicle
  • 60K/day, genome for SARS-CoV-2
  • 200 PB, climate modeling
  • 150 PB, alternative green energy
  • 85 PB, Cancer Genomics
  • 250 billion objects, NLP
Data Sheetdownload
NVIDIA DGX POD™ 최대 레퍼런스 보유1
NVIDIA DGX POD™ 최대 레퍼런스 보유2
NVIDIA DGX POD™ 최대 레퍼런스 보유3

DDN INFO

  • EXA
  • Storage
  • Insight
exa6
exa6

EXAScaler features

  • Hot Pools – Automated data tiering between SSD and HDD pools
  • End-to-end integrity – checksum from client to media
  • Data On Metadata (DoM) – Maximum efficiency for small files *
  • NFS/SMB data services – scalable multi-OS protocol access
  • User analytics – monitor node activity, report file size distribution
  • Multi-tenancy – secure deployment for shared environments
  • Tight GPU integration – container-friendly deployment optimized for GPUDirect, DGX
  • Hot Nodes – Automated read caching to compute-local flash
  • Enhanced Multi Rail – Simplified Client Networking *
  • TRIM – Efficiently secure space and maximize SSD functionality *
  • DDN A3I – a turnkey, fully optimized solution for AI and DL
  • Container-Optimized Client – True Parallel Data Access to Containers
  • DDN DataFlow solution – large-scale backup, archive, movement, synchronization
  • In-Cloud – DDN Parallel File System on AWS, Google Cloud, and Azure

DDN A3I Product Family

ES200NVX2

  • Read 48GB/s, Write 38GB/s
  • 24 NVMe Slots
  • 1.5M IOP/s
  • 4x HDR/HDR100 or 200/100GbE
  • NDR scheduled for release in 2023

ES400NVX2

  • Read 90GB/s, Write 65GB/s
  • 24 NVMe Slots
  • 3M IOP/s
  • 8x HDR/HDR100 or 200/100GbE
  • NDR scheduled for release in 2023
  • Up to 900 x SAS HDD/SSD

SS9024

  • Storage Enclosure
  • 90 Disk Slots (HDD/SSD)
TLC FLASH

Best IOPS & Throughput per rack

  • Up to 70M IOPs in a single rack
  • 1.8TB/s of read throughput
  • 1.4TB/s of write througput
  • 16PB of Flash
QLC FLASH

Best Price per Flash TB

  • Up to 14M IOPs in a single rack
  • Up to 630 GB/s of read throughput
  • Up to 490 GB/s of write througput
  • Up to 26PB of Flash
HYBRID

Best Price per TB

  • 20PB per Rack
  • 90 GB/s of read throughput
  • 65 GB/s of write throughput
  • 1DDN monitoring solution (DDN Insight) provided
  • 2Full system integrated monitoring possible
  • 3Compare and analyze system performance, resources, status, etc.

supported system

SFA storage system, EXAScaler

Minimum Requirements

  • Intel or AMD server: at least 12 cores and 64BG
  • VM: At least 32GB RAM and 8 cores

product features

  • Insight Monitoring UI

    Performance data visualization
    Alerts and event notifications
    Configuration Details
    Email settings

  • Grafana UI

    Ad-hoc querying, analytics, and templating
    Visualize from a variety of sources
    custom layout

  • User Optimized Event Manager

    Event data deduplication
    flapping handler
    concentrated collection
    Group notification emails

  • Scheduler

    Collected from various sources
    Alert and event confirmation schedule
    health check schedule
    Outbound email schedule
    External script integration
arrow_upward
close