About High-Performance Computing (HPC) at Bowdoin

Questions

  • What is High-Performance Computing (HPC) at Bowdoin?
  • What HPC resources are available at Bowdoin College?
  • How do I get started with the Bowdoin HPC cluster?
  • What hardware does the Bowdoin HPC cluster have?
  • Does Bowdoin have GPU computing available?
  • Who can use the Bowdoin HPC cluster?
  • What software is available on the HPC cluster?
  • How do I access HPC resources from off campus?
  • What is the Slurm cluster at Bowdoin?
  • Can I include HPC information in a grant proposal?

Environment

This article applies to Bowdoin faculty, students, and researchers who need access to high-performance computing resources for computationally intensive work. All HPC resources require a Bowdoin account and a connection to the campus network or VPN.

Resolution

What Is the Bowdoin HPC Cluster?

The Bowdoin HPC Cluster is a Linux-based computing environment managed by Bowdoin IT. The cluster uses the Slurm workload manager for job scheduling and runs on Rocky Linux. It is designed for research, coursework, and other tasks that require significant computing power beyond what a desktop or laptop can provide.

Cluster Capabilities

  • Approximately 1,400 CPU cores spread across multiple compute nodes
  • Compute nodes ranging from 16 to 192 CPU cores and 192 GB to 2 TB of RAM
  • Approximately 20 NVIDIA GPU cards including RTX 3080, RTX 2080 Ti, RTX 5090, A100, and Pro 6000 models
  • Dedicated high-speed Gluster parallel filesystem for temporary scratch storage
  • 2x100 GB low-latency Ethernet networking (200 GB per node)
  • Support for single-threaded, SMP (shared memory), and OpenMPI parallel processing
  • A wide range of commercial and open-source scientific software
Grant proposals: Bowdoin College provides a Linux HPC Cluster consisting of approximately 1,400 CPU cores spread across multiple compute nodes, including approximately 20 NVIDIA GPU cards. Compute nodes range from 16 to 192 CPU cores and 192 GB to 2 TB of RAM. The HPC Cluster utilizes a dedicated, redundantly configured Gluster high-speed networked filesystem. Core networking is 2x100 GB low-latency Ethernet (200 GB per node). The HPC Cluster supports single-threaded, SMP, and parallel environments and a range of commercial and open-source software applications.

How to Access the HPC Environment

There are four ways to access HPC resources at Bowdoin:

  • HPC Web Portal at hpcweb.bowdoin.edu — a browser-based interface for shell access, graphical applications, file management, and job monitoring. See Use the HPC Web Portal (Open OnDemand) in the Related Articles section.
  • SSH to moosehead.bowdoin.edu — traditional command-line access to the Slurm cluster headnode. See Access the Bowdoin HPC Environment in the Related Articles section.
  • JupyterLab at jupyter.bowdoin.edu — web-based Python and R notebook environment running on the HPC cluster.
  • RStudio at rstudio.bowdoin.edu — web-based R development environment running on the HPC cluster.
Off-campus access: All HPC resources require a VPN connection when accessing from off campus. See VPN at Bowdoin in the Related Articles section for VPN setup instructions.

Getting Started in 7 Steps

  1. Open Firefox or Chrome and go to hpcweb.bowdoin.edu. Log in with your Bowdoin account name (for example, jsmith, not jsmith@bowdoin.edu).
  2. Upload any data files you need to your home directory or your HPC Research space using the Web Portal file browser. Avoid spaces in file and directory names — use dashes or underscores instead.
  3. Open a command shell on the Slurm headnode by clicking Clusters at the top of the Web Portal and selecting Bowdoin Slurm HPC Cluster Shell Access.
  4. Change into the directory containing your data files using the cd command. For example, cd mydirectory for a subdirectory in your home folder, or cd /mnt/research/jdoe for a research space.
  5. Submit your job to the HPC cluster. See Submit and Manage Jobs on the HPC Slurm Cluster in the Related Articles section for job submission details.
  6. You will receive an email when your job starts running and another when it finishes.
  7. Return to your directory to find the output files containing your job results.

Additional Help

If you need further assistance, you have several options:

  • Bowdoin Bot: Chat with Bowdoin Bot directly from any KB page for instant answers.
  • Phone: Call the Bowdoin College Service Desk at (207) 725-3030.
  • In person: Visit the Tech Hub in Smith Union during business hours.
  • Submit a ticket: Request assistance through the Service Catalog.

Additional Resources

Print Article

Related Articles (6)

Access the Bowdoin HPC Environment
Instructions for connecting to the Bowdoin HPC environment using SSH, the HPC Web Portal, JupyterLab, or RStudio. Covers SSH access from macOS and Linux, VPN requirements for off-campus use, and SSH configuration tips for dropped connections.
HPC Cluster Queues, Policies, and Hardware Overview
Reference information for the Bowdoin HPC Slurm cluster, including queue (partition) descriptions, job policies and resource limits, and a hardware overview suitable for grant proposals.
Submit and Manage Jobs on the HPC Slurm Cluster
Instructions for submitting, monitoring, and managing jobs on the Bowdoin HPC Slurm cluster. Covers writing job scripts, using sbatch and the hpcsub wrapper, running parallel processing jobs (SMP and OpenMPI), running interactive jobs, and controlling jobs with squeue and scancel.
Transfer Files to and from the HPC Environment
Instructions for transferring files between your local computer and the Bowdoin HPC environment. Covers the HPC Web Portal file browser, mounting the HPC research space via SMB from macOS or Windows, SFTP from the command line, and using Gluster temporary scratch storage for running jobs
Use GPU and High-Memory Resources on the HPC Cluster
Instructions for requesting GPU computing, high-memory nodes, and other specialized resources on the Bowdoin HPC Slurm cluster. Covers available NVIDIA GPU cards and request syntax, memory reservation options, mixed GPU and CPU jobs, and the experimental NVIDIA Grace Hopper system.
Use the HPC Web Portal (Open OnDemand)
The Bowdoin HPC Web Portal (Open OnDemand) provides browser-based access to the HPC environment for command-line sessions, graphical applications, file management, and job monitoring. The portal is accessed at hpcweb.bowdoin.edu using Firefox or Chrome.