Neuroimaging does not fit on a laptop. A single fMRIPrep run is hours of CPU and tens of gigabytes; a deep learning model wants a GPU; and datasets like UK Biobank are large enough that where the data lives determines where you compute.
Start the account process early — it is independent of data access, and both take time.
The national cluster¶
Digital Research Alliance of Canada (DRAC) — formerly Compute Canada, and where most of this program’s heavy computation happens. Covers account setup, connecting, tying your work to a sponsoring PI, and long-term storage. Read this one properly; the account and sponsorship steps have real lead time
Running HPC jobs — the scheduler, job scripts, and how to ask for resources without either wasting an allocation or having jobs killed
Performant computing — a short note on making things faster
Interactive work¶
Compute Canada JupyterHub — notebooks on cluster hardware, which is the fastest route from nothing to a running analysis
Platform-hosted alternatives¶
Sometimes the data does not move and you compute where it lives:
UK Biobank RAP — the required route for UKBB
brainlife.io — browser-based pipelines, no cluster account needed
Before you scale up¶
Whatever you run on the cluster has to be reproducible there, which means sorting out dependencies before you are debugging them over SSH at scale. See code environments and containers — covered properly in step 6, but worth a look now if you are about to submit a thousand jobs.
Next: Organize your data.