SC17 Denver, CO

HPC Carpentry - Practical, Hands-On HPC Training

Authors: Andrew Turner (University of Edinburgh)

Abstract: Carpentry training has proven hugely successful; providing researchers with the tools required to exploit software and data to improve their research. The increasing role HPC is playing in research means that more researchers need HPC skills in order to progress. We are designing an HPC Carpentry course to bring the benefit of the Carpentry approach to current and potential HPC users. We will describe the state of HPC Carpentry and facilitate discussions on its Goals, Design, Structure and Implementation. We will use this BoF to employ the expertise and experience of the HPC community to deliver the best course possible.

Long Description: As more and more research comes to depend on the use of high performance computing to progress, there is a clear need to equip researchers with the basic, practical skills so they can exploit HPC resources in an effective way. These skills are particularly important for people just getting started with high throughput and high performance computing, as they make researchers more effective and make the most efficient use of resources. Given the success of the Carpentry series in equipping researchers with the basic, practical skills required to manage and develop their software and scripts (Software Carpentry[1]) and data analysis (Data Carpentry[2]) it is natural to take the Carpentry approach and apply it to HPC.

In this BOF we will gather input from the experts in the HPC community on the Goals, Design, Structure and Implementation of HPC Carpentry. The session will open with a series of short (5 min) presentations from the BOF session leaders setting the context for HPC Carpentry and the thinking that has already taken place. After the introduction and stage-setting we will run a facilitated discussion covering the following topics:

* Goals + Who are the target audiences for HPC Carpentry? + What should the learning outcomes be? * Design + What lessons should be included? + How much modularity do we want? + How (if any) much site/facility specific material would be needed? * Structure + How long should the course last? Should we support different lengths? + Balance of practical, side-along training to lectures * Implementation + Technology for practical work: specific facilities, local VMs, public cloud? + Contribution methodology for any site/facility specific lessons

The ideas and discussion from this BOF will be used to produce a public report describing the key areas of development for HPC Carpentry and this will provide the base for the development of the first version of the training. In addition, all BOF attendees will be invited to join the HPC Carpentry discussion forum on Google Groups to continue providing input to development of HPC Carpentry.

We plan to run the first instances of HPC Carpentry before SC'18 and will submit a follow-on BOF to continue the community development of HPC Carpentry by sharing experiences from the initial events.

[1] [2]

Conference Presentation: pdf

Birds of a Feather Index