About these tutorials

Video

Watch this in the Winter Kickstart 2021 course.

The “Shell crash course” is a useful prerequisite. You can watch a short version (20 min) or longer version (1 hour).

Or see the full playlist.

Welcome to the Aalto Scientific Computing High-performance computing (HPC) tutorials. These tutorials will get you started with the Triton cluster.

Despite the HPC in the name, most of these tutorials are not about the high-performance part: instead, we get you started using and submitting jobs to the cluster. These days, many people use a cluster for simple jobs: getting more stuff done at once, not a few very big tasks. Doing the big tasks are a more specialized topic, which this will introduce you to and you will be able to use other software for that. Programming your own HPC software is out of our scope.

Not at Aalto?

These tutorials use Aalto’s cluster as an example, but they are designed to be as general as possible. You can probably follow along and learn useful topics here anyway, but you will have. We recommend you review the basics of your cluster’s documentation first, then read this. You will have to translate somethings to your own specifics, but hopefully our examples can inspire you to understand your own docs better.

These tutorials will be quite useful if you have:

  • A cluster using Slurm as the batch system

  • You can get a shell on that server

  • The git installed

  • Python 2 or 3 installed as python

Unfortunately, not all clusters are standardized (though probably they should be slightly more than they are). Things that may be different (we’ll try to point these out where relevant):

  • The way you connect to the cluster, including remote access methods.

  • Exact names of batch partitions.

  • The slurm utility probably isn’t installed, seff may not be there.

  • Module names for software.

  • You probably don’t have our singularity stuff installed.

  • Parallel and GPU stuff is probably different.

Runing another cluster?

If you run your own cluster, create a quick reference such as Triton quick reference so that others following tutorials such as this can quickly translate to your own cluster’s specifics.

Don’t you think clusters should be more interoperable? It’s something we are thinking about.

What’s next?

Introduce yourself to the cluster resources at Aalto.