Running a Workshop
If you have an audience that would benefit from an HPC Carpentry workshop, this page will provide guidance about running a workshop! To help us help you, please do the following before getting started:
Identify your audience
HPC Carpentry materials are aimed at a research audience who is new to the world of
remote, large-scale computing. See Why HPC Carpentry for
three “learner profiles” of researchers who would benefit from an HPC Carpentry workshop.
Does one of these profiles resonate with you? If not, think about the audience for your own workshop. What skills are they likely to have (or not have) before the workshop. What is their motivation and what will be most immediately useful for them to know?
Choose workshop lessons
Based on your audience (and desired workshop length/format), choose the relevant workshop lessons for your workshop. A standard two day HPC Carpentry workshop looks like this:
- Day 1: HPC Introduction/Shell Skills
- Day 2: Python pipelines OR Introduction to Chapel
For Day 1, the workshop begins with the first episode of
hpc-intro and then can
incorporate shell lesson material of your choice (HPC-shell, SWC shell, etc.) before
going through the rest of the intro material.
Note: It is strongly advised to include an introduction to the shell, especially for a novice audience. It is extremely difficult to ensure that all participants have the necessary shell skills in advance, so incorporating them into the workshop is usually the best way to go.
Some additional customized examples are listed below.
As an instructor, you should ideally set up or arrange access to a computing cluster for your workshop. Each student will need their own account on the cluster. Our materials assume that you are using some form of environment modules to manage software installation, and use a batch scheduler to schedule jobs. Though our materials have been developed for systems using the SLURM scheduler, it should be fairly easy to adapt them to other systems - the concepts covered are not scheduler-specific.
In terms of venue, you should arrange access to a room with wifi, a projector, and enough power outlets or extension cords for students to bring their laptops. You can also quickly create a lesson website using the Software Carpentry workshop-template and arrange for attendee registration through Eventbrite.
If you want to customize the materials, we recommend “forking” them to a Github account
associated with you. The
fork button at the top right
of each main repository page will prompt you to create a copy of the materials
in your own or an organizational account. There
will be a corresponding website with these materials at the link:
Once this copy is created, you can edit
the lesson materials (contained in the
_episodes folder) directly on Github or, if
you’re familiar with remote git workflows, through a cloned copy of the fork.
In addition to the steps above, there are several additional pieces of setup to be performed depending on which lessons are being taught.
Analysis Pipelines with Python
Although most work will be performed on student’s laptops, the final segment (scaling jobs across a cluster) will occur on your cluster environment. You should have some version of Python 3 preinstalled for students to use. Anaconda Python is a good choice for most use cases. Don’t preinstall Snakemake on the system for users (unless it’s already there), as this lets students practice installing local Python packages on their account.
Parallel Computing with Chapel
Chapel can be a tricky package to setup for newcomers and does not run on Windows (except through something like Docker or Cygwin). We recommend setting up an installation of Chapel on your cluster beforehand for use in the workshop. For instructions on compiling/installing Chapel, see the Chapel quickstart documentation, specifically the sections on “Using a more full-featured Chapel” and “Using Chapel in multi-locale mode”. It is also possible to use the Chapel Docker image for teaching on student laptops. If you choose to use the Chapel Docker image, make sure students install Docker before the workshop.