Your Browser is Out of Date

Nytro.ai uses technology that works best in other browsers.
For a full experience use one of the browsers below
Scalability Guidelines for Deep Learning with High-Speed Object Storage
Executive summary Introduction Solution architecture
Overview
Storage – Dell EXF900 object storage
Dell PowerEdge servers
PowerSwitch Networking
Bill of materials
Deep learning training performance and analysis Benchmark setup Conclusion References

Bill of materials

Home > Storage > ObjectScale and ECS > Industry Solutions and Verticals > Scalability Guidelines for Deep Learning with High-Speed Object Storage > Bill of materials

Bill of materials

  • Table 1.      Bill of materials

    Component

    Purpose

    Quantity

    Dell EXF900

    • 46 TB (12 x 3.84 TB NVMe Drives)
    • 192 GB RAM
    • Two 25 GbE interfaces

    Shared storage

    5 nodes

    Dell PowerEdge R7525

    • 2 x AMD EPYC 7H12 CPU @ 2.6GHz with 64 Cores /128 Threads).
    • 512 GB RAM
    • BOSS-S2 controller card with 1x M.2 240GB Drive
    • 1x 960 GB SSD Drive (unused)
    • 2x NVIDIA A100 GPU 80GB PCIe
    • 2x 100 GbE Network (Melanox MT2892 Family [ConnectX-6])
    • 2x BCM57414 NetXtreme-E 10Gb/25Gb RDMA Ethernet Controller
    • 2x NetXtreme BCM5720 2-port Gigabit Ethernet PCIe

    Compute Server

    8

    Table 2.      Software versions that were tested for this document

    Component

    Version

    Dell EXF900

    3.8.0.1

    Dell R7525 – Linux kernel

    5.4.0-150-generic

    Dell R7525 – Ubuntu

    20.04.6 LTS

    NVIDIA CUDA

    12.1

    NVIDIA Driver

    530.30.02

    NVIDIA NGC TensorFlow Image

    nvcr.io/nvidia/tensorflow:20.12-tf2-py3

    TensorFlow Benchmark

    https://github.com/tensorflow/benchmarks