Introduction

When running a workshop to teach learners how to use an HPC, an instructor is immediately presented with a few problems:

1. Very few users ever get to see an HPC in real life and it is left to imaginations and sci-fi movies to visualise what an HPC is. To many this is quite a scary concept.
2. Training on a “real” HPC can cause learners to be anxious that they might “break” something.
3. Access to an HPC needs to be arranged. This is not always a trivial task as the use of HPC resources can be quite restricted in terms of who are allowed to use a specific HPC.
4. Workshop attendees often do not read their emails requesting them to create accounts before they turn up for the workshop which results in instructors having to create accounts on the day. Apart from quite often delaying the start of the workshop, it is also not always possible for instructors to create the user accounts on the day.
5. HPC resources are always in demand and running a workshop on a “real” HPC takes resources away from “real” processes running at the time.
6. HPCs typically have to be connected to via the Internet. Any issues with accessing the Internet will affect the workshop.
7. If an HPC is heavily used or if someone runs a job on the login node, learners might not be able to log in or there are significant delays in getting jobs into queues which again affects the timing of the workshop.

All these mentioned issues (and probably more) can be addressed by having a dedicated HPC for training. But usually “real” HPCs are very expensive and it wouldn’t be feasible to purchase typical high-end HPC hardware just for a training setup. However, it is completely possible to use low-end hardware to create a cluster that will run almost all the required software to learn how to use an HPC.