Getting Started with Open OnDemand¶
What we're cooking
How to access Unity through a browser using Open OnDemand, and when to choose the web portal over the command line.
What is Open OnDemand?¶
Open OnDemand (OOD) is a web portal that lets you use the HPC cluster entirely through your browser. No SSH client, no terminal setup, no command-line knowledge required to get started. Once you log in, you have access to:
- A file browser for navigating, uploading, and downloading files
- A job management interface for creating, templating, and monitoring Slurm jobs
- Shell access (a full terminal in your browser tab)
- Interactive apps (launch JupyterLab, RStudio, VS Code, MATLAB, Mathematica, and more with a few clicks)
- A Module Explorer for browsing available software without the command line
OOD is designed to lower the barrier to entry for HPC. If you've been putting off learning the command line, OOD lets you start doing useful work on the cluster right away. That said, OOD and the command line aren't mutually exclusive; most researchers end up using both.
New to the cluster?
If this is your first time accessing Unity, start with Getting Started to create your account and understand the cluster architecture. OOD and SSH are two different ways in; the cluster itself is the same either way.
When to use Open OnDemand vs. SSH¶
Choosing between OOD and the command line isn't an either/or decision. The table below outlines where each approach shines.
| Situation | OOD | SSH + CLI |
|---|---|---|
| New to Linux / command line | ✅ Better | ❌ Steep learning curve |
| Uploading/downloading a few files | ✅ Drag and drop | Works fine |
| Running Jupyter Notebooks | ✅ One click | Requires SSH tunneling |
| Running RStudio | ✅ One click | Requires X11 or tunneling |
| Automating repetitive jobs | ❌ Manual only | ✅ Scripts and cron |
| Complex job pipelines | ❌ Limited | ✅ Full control |
| Transferring large datasets (>1 GB) | ❌ Slow | ✅ rsync / Globus |
| Working offline / scripting | ❌ Browser required | ✅ Works anywhere |
OOD and SSH complement each other. Many researchers use OOD for exploration and prototyping, then switch to job scripts for production runs. They share the same cluster, the same storage, and the same files; switching between them is straightforward.
Accessing Open OnDemand¶
-
Open a browser and go to https://ood.unity.rc.umass.edu/.
-
You'll see a federated login page listing participating institutions. Select Dartmouth College from the dropdown, then log in with your campus credentials (the same username and password you use for other Dartmouth services).
-
After login, you'll land on the OOD dashboard.
New accounts take a few minutes
If you just created your Unity account, it may take up to 10 minutes before you can log in to OOD. If you see an authentication error immediately after account creation, wait a few minutes and try again.
The dashboard¶
The OOD dashboard is organized around a top navigation bar. Here's what each menu does:
- Files — Opens a file browser for your home directory and other storage locations. You can upload, download, create, and delete files here.
- Jobs — Contains three tools:
- Job Composer for creating and submitting batch jobs
- Templates for saving and reusing job configurations
- Active Jobs for monitoring running and queued jobs (based on
squeue, so completed jobs disappear after about 5 minutes)
- Shell — Opens a terminal that connects you to Unity exactly as SSH would, but running in your browser.
- Interactive Apps — Launch graphical applications on compute nodes: JupyterLab, RStudio, VS Code, MATLAB, Mathematica, Virtual Desktop (XFCE), and more. This menu also includes the Module Explorer, a searchable web interface for browsing available software (equivalent to
module spideron the command line).
Managing files¶
The Files menu opens a file browser that works much like a desktop file manager. From here you can:
- Navigate your home directory, work space, and scratch space using the path bar at the top
- Upload files by dragging them from your desktop and dropping them into the browser window, or by clicking the Upload button
- Download files by selecting them and clicking the Download button
- Create, rename, and delete files and directories using the toolbar
- Edit text files directly in the browser
- Open in Terminal to jump from a directory straight into a shell session
OOD file transfer has a size limit
The OOD file browser is convenient for small files but slow and unreliable for large ones. For anything over ~100 MB, use rsync or Globus instead. Uploading large datasets through the browser risks timeouts and incomplete transfers.
Shell access¶
The Shell menu → Unity Shell Access opens a terminal in your browser tab. This is functionally identical to an SSH session:
- Same files and directories
- Same environment modules
- Same job submission commands (
sbatch,squeue,scancel) - Same resource limits
Shell access through OOD is particularly handy when you're on a managed computer where installing an SSH client is restricted, or when you just want a quick terminal without leaving your browser.
Tip
Everything you can do over SSH (loading modules, submitting jobs, editing files with vim or nano, running scripts) works exactly the same in the OOD shell.
Avoid conda init in the OOD shell
If you have a self-installed version of Anaconda or Miniconda and have run conda init, the OOD shell may crash on startup. Use the pre-installed Unity modules (module load conda/latest) instead, and avoid running conda init. If your shell is already broken, remove the conda initialization block from your ~/.bashrc to fix it.
What's next¶
- Interactive Apps in Open OnDemand — launch Jupyter, RStudio, VS Code, and more through OOD without any SSH tunneling
- Submit Your First Job — running batch jobs via the Job Composer or the command line
- Transferring Data — efficient file transfer for larger datasets using rsync and Globus