NETL's private Infrastructure as a Service (IaaS) cloud computing environment

Part of the Center for Artificial Intelligence and Machine Learning (CAML)


Accelerate deep learning and training with the power of GPU tensor cores


Access from anywhere — from home, the office or on the road

Big Data

Store even the largest of datasets


Watt is comprised of the OpenStack software platform written in Python running on modern, bare metal, GPU accelerated servers.

CAML Hardware

  • 24 40-core Intel Skylake nodes
    • 384 GBs of RAM
    • Four Nvidia P100 GPUs
  • 40 PBs of raw storage
    • Configured as Lustre parallel file system, XFS, and ZFS shares
  • 100 Gbps OmniPath interconnect for parallel file system and high performance file access
  • 50 TFLOPs performance across all of the CPUs
  • 323 TFLOPs performance across all of the GPUs
  • 66 GB/sec write and 122 GB/sec read performance on the parallel FS

OpenStack Software

  • OpenStack is installed and used to manage all ML compute and
    storage resources
  • Unlike traditional HPC environments where OS versions and libraries
    are often standardized and difficult to reconfigure, OpenStack allows
    users to “bring their own” OS images and environments
  • Can use either Windows or Linux, and any Linux variant can be
  • We provide baseline Windows and Linux image files to get you started
  • Can work with you to build OS images tailored to you specific
    application requirements

A Look at Watt

Watt (the SI unit of power) is named after Scottish inventor, mechanical engineer, and chemist James Watt

Get Started

Please email to request access to the CAML system. Provide whatever information you can about your intended application suite, work flow, and data storage requirements. We will work with you to develop storage plans, VM templates, and allocation project resources to meet your needs.

Download the CAML Client

Download the CAML Access Client for your platform using the following link(s):

Download Documentation

Learn how to launch and use a new instance, create a new virtual network, create and attach a new volume to an instance, and create a share network.
Scroll to Top