Welcome to Svante user documention

Acknowledgement Statement:

Please acknowledge the use of the Svante HPC cluster in publications as:

The computations presented here were conducted using the “Svante” cluster, a facility located at MIT’s Massachusetts Green High Performance Computing Center and supported by the Center for Sustainability Science and Strategy (https://cs3.mit.edu).

Contents:

• 1. Who to Contact
• 2. General Information
  • 2.1. How to log into Svante
  • 2.2. /home spaces
    • 2.2.1. Public space subdirectory
  • 2.3. File servers
  • 2.4. Compute nodes
  • 2.5. Archive space
  • 2.6. Running background screen sessions
  • 2.7. Svante node specs
• 3. Module System
• 4. Using SLURM to Submit Jobs
  • 4.1. SLURM batch script example
  • 4.2. Common #SBATCH options
  • 4.3. How to submit a SLURM job
  • 4.4. Requesting resources for multi-core, multi-node jobs
  • 4.5. Interactive SLURM sessions
  • 4.6. Further SLURM script examples
• 5. Python and Jupyter Notebooks (svante-ood.mit.edu)
  • 5.1. Running Python on Svante
    • 5.1.1. Terminal-window based Python with pre-installed packages
    • 5.1.2. Installing new packages in Python
  • 5.2. Running Jupyter Notebooks using Svante Open OnDemand
• 6. Best Practices for Svante Use
  • 6.1. Use of login node (aka “head node”) vs. file servers vs. compute nodes
    • 6.1.1. Login node usage
    • 6.1.2. File server usage
    • 6.1.3. Compute node usage
  • 6.2. Where should I compile code for running on Svante?
  • 6.3. Which compute nodes should I use: HDR, HDR200, EDR, or FDR?
  • 6.4. Where should I write model output?
  • 6.5. Limits on compute usage
  • 6.6. Reservations
  • 6.7. Courtesy and common sense
  • 6.8. AI agents and coding assistants