Getting Started with ARC

  1. Connecting to an ARC cluster
    1. Step-by-step Connection Guide
      1. Trouble Shooting
  2. Transferring Data
    1. Using Command Line Tools (scp)
    2. Using Globus
      1. Logging Into Globus
      2. Transferring Files
      3. File Transfer Options
  3. Using Python
    1. Setting Up a Python Virtual Environment
    2. Installing Python Packages
      1. Recommended Method
      2. Alternative Methods
    3. Pre-installed Packages
  4. Software Available

Connecting to an ARC cluster

ARC systems are remote, meaning you must connect to them from your local machine. This is typically done using a secure shell (SSH) connection.

Before connecting, ensure you have:

  • A terminal or SSH client installed (e.g., Terminal on macOS/Linux, PuTTY on Windows)
  • Your CCDB system username and password

Step-by-step Connection Guide

  1. Open Your Terminal: On macOS or Linux, open the Terminal. On Windows, use Command Prompt, PowerShell, or an SSH client like PuTTY.

  2. Connect to the ARC cluster: To connect to the Cedar supercomputer, type the following command:

     ssh USERNAME@cedar.alliancecan.ca
    • Replace USERNAME with your CCDB username.
    • To log into another ARC cluster, replace cedar with its name (narval, beluga, etc.)

  3. Enter your password: If you are not using SSH key authentication, you will be prompted to enter your password.

  4. Authenticate using MFA: Follow the multi-factor authentication prompt (e.g.: approve the request on Duo).
⚠️ Important: Once you are logged in, you will be in the HOME directory.

Learn more about ARC storage locations.

Trouble Shooting

  • If the SSH connection fails, ensure your internet connection is stable.
  • Verify that your username and password are correct.
  • If using SSH keys, confirm they are correctly configured.

Transferring Data

Tranferring data between your local computer and remote clusters is an important part of working with HPC systems. The method you choose depends on the size of the data, the systems involved, and the specific requirements of your project.

Login nodes have internet access, so commands such as curl or wget can be used to download files directly from the internet. The scp command is ideal for simple transfers between your personal computer and remote ARC clusters. For larger datasets or transfers between two ARC clusters, Globus provides high-speed, secure transfers, making it the best option for large-scale file transfers.

Using Command Line Tools (scp)

The scp command is used to securely copy files or directories between a local (your personal computer) and a remote (HPC cluster) host, or between two remote hosts. It uses SSH for data transfer and provides a secure way to transfer files over the network.

scp <source> <destination>

Copying a file from local to remote

scp myfile.txt user@remote_host:/path/to/remote/directory/
  • myfile.txt: The file you want to transfer from your local system.
  • user@remote_host: Replace user with your username on the remote system, and remote_host with the remote server’s address (it could be an IP address or domain name).
  • /path/to/remote/directory/: The directory on the remote server where the file will be placed.

Copying a file from remote to local

scp user@remote_host:/path/to/remote/file.txt /path/to/local/directory/

Copying a directory from local to remote

scp -r my_directory/ user@remote_host:/path/to/remote/directory/

The -r flag (stands for recursive) tells scp to copy the entire directory and all its contents.

Using Globus

Globus is a fast, reliable, and user-friendly tool for secure file transfers. It allows you to easily move files between your personal computer and an ARC cluster or between two ARC clusters.

In Globus, a collection is a designated storage location that allows you to transfer and manage files. Collections can be hosted on HPC clusters, cloud storage systems, or personal computers, and provide a seamless way to move data across different systems.

⚠️ Important: To transfer files between your personal computer and ARC clusters using Globus, you must install and run Globus Connect Personal. This allows your computer to act as a Globus collection for secure data transfers.

Logging Into Globus

  1. Go to the Alliance’s Globus Portal.

    In the Use your existing organizational login box, select Digital Research Alliance of Canada.

    globus

    Press Continue.

  2. Log in using your CCDB username (not your email) and password.

  3. Open File Manager in the Globus portal.

    globus_dash

Transferring Files

  1. Use the Search bar to find collections by name.

    For example to transfer files to or from Narval, type narval and select the Compute Canada - Narval option. globus_narval

  2. Once the collection is selected, you will see a ist of directories and files. Navigate them by double-clicking the folder or using the up one folder button. globus_narval_dirs

  3. To start a transfer, select a file or directory by single-clicking. To select multiple files or directories, hold Ctrl.

  4. Press the blue Start button to start the transfer. globus_narval2beluga

  5. Globus will assign a unique job ID, and the transfer will begin immediately.

  6. To track the progress of your file transfer or view past transfers, click on Activity in the left sidebar. globus_narval_activity

    Once your transfer is complete, you will receive an email. globus_email

File Transfer Options

Globus also offers a variety of file transfer options through the Transfer & Sync Options button. These allow you to:
✅ Sync transfers (only move new or modified files)
✅ Delete destination files that don’t exist in the source
✅ Preserve file modification timestamps
✅ Verify file integrity after transfer (enabled by default)
✅ Encrypt the transfer (note: encryption is slow and should only be used only for sensitive)

globus_file_options

For more details on using Globus, refer to the Alliance’s Globus Documentation page.

Using Python

When working with Python on ARC clusters, it’s best to set up a virtual environment (venv). This helps keep your Python code runs with the correct dependencies, no matter which compute node processes your job.

ARC clusters use wheels which are pre-compiled Python packages that are optimized or the clusters and help to speed up installation and reduce disk usage.

Setting Up a Python Virtual Environment

  1. Check which versions of Python are available. 
     module avail python
  2. Load the desired Python version. 
     module load python/3.10
  3. Create a virtual environment. 
     virtualenv --no-download ENV

    This command creates a new virtual environment called ENV. The --no-download flag ensures that the wheels optimized for the ARC clusters are used.

  4. Activate the virtual environment. 
     source ENV/bin/activate
  5. Upgrade pip. 
     pip install --no-index --upgrade pip

    This command updates pip to the latest version and the --no-index flag ensures that the ARC-optimized wheels are used.

  6. Exit the virtual environment. 
     deactivate
⚠️ Important: While it can be a great option for managing packages on your local computer, do not use Anaconda on ARC clusters.

Installing Python Packages

To install Python packages in your virtual environment, the best approach is to use wheels, which are optimized for the ARC clusters. This makes the installation process much faster and more efficient.

Install packages using pip within your job script.

pip install --no-index <package_name>

The --no-index flag ensures that pip uses wheels, which are pre-compiled and tailored for ARC systems. This makes installation faster and helps conserve disk space in your /HOME directory.

Alternative Methods

If wheels are not provided for your Python package, you can install packages from GitHub or use the centralized web-based version of pip on the login node.

⚠️ Important: Installing packages using GitHub or the web-based `pip` version is not ideal for frequent or large installations.

When using these alternative installation methods, in your job script, you need to create and activate your virtual environment in your \HOME directory:

source $HOME/ENV/bin/activate

Pre-installed Packages

ARC clusters come with a variety of pre-built Python wheels that are optimized for the cluster’s architecture, allowing for fast and efficient installations. These include popular libraries commonly used in scientific computing, machine learning, and bioinformatics. Some of the pre-installed packages include:

  • PyTorch
  • Biopython
  • Scikit-learn (sklearn)
  • Many more!

For convenience, the SciPy stack is available as a module:

  • NumPy
  • SciPy
  • Matplotlib
  • IPython
  • pandas
  • Sympy
  • nose

To load the SciPy stack, use:

module load scipy-stack
💡 Important: In addition to Python, ARC clusters also support other programming languages, including R, C++, MATLAB, Java, etc.

Software Available

ARC clusters also have a variety of scientific software pre-installed from various disciplines. Some examples include:

Discipline Software
🔬 Chemistry RDKit, CP2K, LAMMPS, GROMACS, ABINIT, AMBER, Atomicrex
🧬 Bioinformatics Plink, REGENIE, Rosetta, IQ-Tree, Bowtie-2, Samtools, MUSCLE
⚙️ Physics / Engineering Quantum ESPRESSO, Abaqus, OpenFOAM, MESA, STAR-CCM
🌍 Earth Sciences CCSM, CDO, GRASS GIS, PROJ, QGIS

For a complete list of available software, visit the Alliance’s Documentation.