Repositories discussion

Repositories Discussion

September 12, 2016

Collections of Training Materials

The discussion began around how to find good training resources. There are numerous repositories, and we would benefit from making it easy for people to be able to search across all of the repositories to find the training materials they want/need.

People would like to get the following from this effort:

- Gather all resources and rearrange them so that they align with the training program on our campus.

- easily searchable, perhaps ways to comment and/or "thumbs up" to show that it's good (crowd sourced curation)

- a shared set of categories/tags/descriptions

There is a range of interest in topics from basic skills, to developing parallel codes, to having access to example codes.

Please share your links so they may be included on the IHPCTC google site:

https://sites.google.com/a/lbl.gov/hpc-training-best-practices/training-resources

XSEDE is developing a process to post links to collections of training materials. All of the links provided to this group will be included. The links will also be included in the HPC University portal - http://hpcuniversity.org

Searches seem to be inefficient for someone new to an area to find topics of interest.

The following links were provided (and will be posted to the above sites):

- Also working on converting Henry's SCIPE (supercomputing in plain English) to a github thing.

o http://www.oscer.ou.edu/education.php

- https://github.com/OSU-HPCC/example_submission_scripts

- http://mpitutorial.com/

- https://www.csc.fi/hu/web/research/-/prace-launches-codevault-repository-for-hpc-code-samples

- http://www.training.prace-ri.eu/nc/training_courses/index.html

- http://hpcuniversity.org/

- https://portal.xsede.org/training/course-catalog

- https://github.com/OSU-HPCC/hpc-novice-first-draft/blob/master/lessons/01-what-is-hpc.md

Glossary of HPC Terms

A glossary of terms would be useful

- some existing glossaries include

o https://www.nics.tennessee.edu/hpc-glossary

o https://cvw.cac.cornell.edu/main/glossary

o https://wikis.nyu.edu/display/NYUHPC/Glossary

o http://wotug.org/parallel/acronyms/hpccgloss/all.html

o https://www.ibm.com/support/knowledgecenter/SSENRW_4.1.1/glossary_hpc/glossary.html

o http://www.intel.com/content/www/us/en/support/topics/glossary.html

o http://www.hpcwales.co.uk/glossary

o http://www.sdsc.edu/news_and_events/glossary.html

If you know of other good glossaries, please share. These will be posted to the IHPCTC google site.

HPC Carpentry

There was discussion about HPC Carpentry. There is a clear consensus that Software and Data Carpentry are a good approach to scaling up training on these topics.

A complementary effort has begun for HPC training, though it is moving slowly. There are multiple views of what should be included that require further discussion. Through the polling tool, it seemed there were multiple people interested in having further discussions on this topic.

A doodle will be issued to find a time to discuss this further.

Code Repository

- There is strong interest in developing a shared code repository.

- Perhaps use github to collect these materials.

- A mechanism for check-in and approval would be useful.

- Use mechanisms like Software Carpentry uses.

- Create a library of examples.

- Keeping the examples current will be important.

- Need a library of scripts

o For example – matlab scripts

There was discussion about a shared repository for

examples of how to run software on a cluster. Dana Brunson and Bennet Fauber set up an example in GitHub as a first pass at a format that might be useful as a starting point for conversation. Please see:

https://github.com/ihpc-training/application-examples

This was done quickly, to illustrate the idea.

Please UpThumb or DownThumb whether you think this basic structure has enough promise for you -- yes, you -- to expend at least 60 minutes of effort contributing to it.

If you do, and have other comments, please open a separate Issue for the topic.

Please take a look at any other open issues to see if the topic, or something close, exists. The conversations should be clearly delineated, but too many will be unwieldy.

If there is enough interest, a separate call will be arranged.