ERISXdl Getting Started

Posted on February 23, 2021
Updated on

February 1, 2023

ERISXdl Pricing Policy

Please bear in mind that in order to recoup the initial capital cost for the ERISXdl platform a fee for GPU usage will be charged at a rate of $0.008 /min GPU.

However, for the purposes of initial testing and debugging a free tier called the Basic partition will also be made available to registered users and which will permit 10 min single GPU jobs up to a maximum of 20 jobs per month

Connecting to ERISXdl

To access ERISXdl Platform you need to be familiar with Linux and command-line interface and request an account to Scientific Computing (SciC) Linux Clusters, if you already have one you can use the same. You can ssh within the Mass General Brigham network or via VPN This method requires some familiarity with the Linux operating system.

Connect from a PC running Microsoft Windows:

  • (Recommended) Use the MobaXterm application to connect. 
    For more information on the features of MobaXterm, reference their documentation.

    • Download the Free Home Edition of MobaXterm here and extract the files from the download to your preferred location. Run the installer and then application if you downloaded the Installer edition, or run the .exe file if you downloaded the Portable edition.
    • Once the application is open, select "Session" from the top toolbar. This will open a new pop-up window.
    • Select "SSH" as the session type, and fill in the specified fields as followed. Once done, click "OK"

      • Remote host: erisxdl.partners.org
      • Specify username (make sure it's checked off): your Mass General Brigham user ID
      • Port: 22
    • A new terminal tab should open in the MobaXterm window, prompting for your password. Enter your Mass General Brigham password to continue.
    • Once logged in, you will also be able to view your home directory files and folders in the "Sftp" tab on the left toolbar in MobaXterm
  • Use the PuTTY ssh client to connect

    • Download the PuTTY ssh client (putty.exe)
    • Double click the downloaded file
    • Enter the following connection information and press "Open"

      • Host Name (or IP address): erisxdl.partners.org
      • Port: 22
      • Protocol: SSH
    • Enter your Mass General Brigham user ID and password when prompted

Connect from Mac or Linux

  • Open the "Terminal" application
  • Using your Mass General Brigham user ID, type this line followed by your password: ssh <username>@erisxdl.partners.org

Select and Prepare the Application Container 

There are several prepared Docker containers stored in Harbor registries that have been optimized by NVIDIA for the ERISXdl environment. These are located under the project called library. Additionally the user may scan the available catalog of Nvidia images at https://catalog.ngc.nvidia.com/ and pull the latest images directly from nvcr.io/nvidia e.g. nvcr.io/nvidia/tensorflow:22.08-tf2-py3.

To see what images you have loaded in the user directory (your local podman runtime):

$ podman images

 
To use a container, select an application container from the ERISXdl Harbor registry:
  • rapidsai
  • tensorflow
  • pytorch
  • clara

Or bring your own application container. To do this, you may pull containers from another outside registry and push them into your Harbor registry to use them throughout the ERISXdl cluster.

You will need to prepare your container for deployment on ERISXdl and run the container image to be accessible. (This step currently requires working with an admin to prepare and deploy the docker image. We are working to open the Docker deployment for users.) Once ERISXdl is in full production and out of pilot phase, every user who signs up for an ERISXdl account will have to be part of a charge group (PAS) to which GPU compute fees will be charged. The PAS groups can also apply for Harbor projects (free of charge) in which group members will be able to store their containers. This will replace the former arrangement of the pilot phase when users had their own personal Harbor projects. The initial limit of 50GB of Harbor storage will continue to apply until further notice.
 
For more information on using containers in ERISXdl, read the Using Docker Containers article.
 
Once your application container or environment (like Conda) is setup you can prepare the bash script to run it from SLURM. The resource manager is SLURM, similar to LSF on SciC Linux Clusters. To learn how to submit jobs on ERISXdl, please read the Using SLURM article.

If you have any questions, please contact us at hpcsupport@partners.org.

 

 

Go to KB0038524 in the IS Service Desk