|
| 1 | +--- |
| 2 | +layout: page |
| 3 | +title: "The BoF at SC21" |
| 4 | +authors: ["Andrew Reid"] |
| 5 | +date: 2021-12-14 |
| 6 | +time: "00:10:00" |
| 7 | +teaser: "We talked to the world, and it talked back!" |
| 8 | +tags: ["Community", "Lesson development"] |
| 9 | +--- |
| 10 | + |
| 11 | +## The BoF at SC21 |
| 12 | + |
| 13 | +The HPC Carpentry team hosted a "Birds of a Feather" session |
| 14 | +at the [SC21][sc21] conference. This was a hybrid conference, |
| 15 | +and the BoF sessions were done via Zoom, with Q&A done via the |
| 16 | +Sli.do tool. Andrew Reid was on-site and hosted the on-site |
| 17 | +portion from the podium, while Trevor Keller, Annajiat Alim |
| 18 | +Rasel, Alan O'Cais, and Wirawan Purwanto were on-line, |
| 19 | +monitoring the Zoom and Sli.do questions and keeping the |
| 20 | +session flowing for all participants. |
| 21 | + |
| 22 | +The primary purpose of this was to try to reconnect with the |
| 23 | +HPC user community, and gather feedback and guidance on how to |
| 24 | +prioritize our efforts moving forward. The BoF was a success |
| 25 | +on all counts: 20 people attended in-person in St. Louis, and |
| 26 | +40 people attended via Zoom, with good, sometimes vibrant |
| 27 | +participation from both "factions" 😇 in the Q&A. |
| 28 | + |
| 29 | +We imagine that the SC21 audience is probably somewhat more |
| 30 | +performance- and hardware-aware than the Carpentries community |
| 31 | +at large, and possibly also biased towards facility operators, |
| 32 | +rather than novice users. The feedback we gained is clearly |
| 33 | +valuable; nevertheless, we need to keep this intrinsic bias in |
| 34 | +mind while we evaluate and act on this input. |
| 35 | + |
| 36 | +There were a number of survey questions planned in advance, |
| 37 | +which were presented via the polling functionality in the |
| 38 | +Sli.do. This information was captured and is available |
| 39 | +in the [coordination repo][coordrepo]. |
| 40 | +The principal results are that attendees would like to see, |
| 41 | +in descending order of preference: |
| 42 | +Future lessons on containers; Python and MPI; Dask and |
| 43 | +Julia; Ability to mix and match lesson content; |
| 44 | +Constructing half-day lessons that use the templating |
| 45 | +capability to adapting lessons to permanent on-premise |
| 46 | +facilities. |
| 47 | + |
| 48 | +After the planned polling questions were done, a more general |
| 49 | +discussion ensued, largely captured in a [CodiMD][bofcodi] |
| 50 | +document. |
| 51 | + |
| 52 | +A number of attendees were interested in how to give feedback |
| 53 | +on their experience with the lessons, and seemed uncertain what |
| 54 | +feedback we're looking for. The answer is that we are mainly |
| 55 | +looking for actionable feedback¹ that helps improve the lessons |
| 56 | +for the community. The main mechanism by which we imagine this |
| 57 | +happening is through instructor and learner comments via poll |
| 58 | +(to be written), Slack, and GitHub issues. At the end of the |
| 59 | +lesson, following usual Carpentries practice, organizers |
| 60 | +should ask learners for positive and negative feedback, on |
| 61 | +green or red sticky notes, and conduct a post-workshop survey. |
| 62 | +This info, particularly the red stickies, is what we're |
| 63 | +looking for, and ideally would be translated into issues on |
| 64 | +the appropriate lesson repository. |
| 65 | + |
| 66 | +We will be digesting the feedback received over the next few |
| 67 | +weeks, and look forward to working through the [generated |
| 68 | +issues][milestone] over the coming months to ensure our |
| 69 | +lessons benefit from this excellent community participation. |
| 70 | + |
| 71 | +### Highlights |
| 72 | + |
| 73 | +- One way HPC Carpentry can add value, beyond the lessons |
| 74 | + themselves, is as a clearing house for HPC educational |
| 75 | + resources generally. |
| 76 | +- The lessons and resources we choose to focus on are a |
| 77 | + signal to the community about our priorities, so we |
| 78 | + should take some care in selecting these. |
| 79 | +- MPI, implemented with low-level programming languages |
| 80 | + (C, C++, Fortran), is still the most commonly-used framework |
| 81 | + for parallel programming. It's appropriate for |
| 82 | + HPC Carpentry to develop lessons that focus on this |
| 83 | + foundational material. |
| 84 | +- New users are increasingly unfamiliar with the command line |
| 85 | + and file system hierarchy. Extra instructional effort may be |
| 86 | + needed to help bridge this gap, including the effort of |
| 87 | + developing GUI-focused lessons based on Open OnDemand |
| 88 | + or JupyterHub. |
| 89 | +- Some users still struggle with the difference in character |
| 90 | + between distributed and shared HPC resources, and more familiar |
| 91 | + laptop or workstation resources. HPC is not just a big |
| 92 | + laptop. |
| 93 | +- Interest in cloud-based HPC resources is surprisingly high, |
| 94 | + driven by people with possibly-transient HPC workloads but without |
| 95 | + institutional HPC facilities. |
| 96 | + |
| 97 | +--- |
| 98 | + |
| 99 | +1. "Actionable feedback" may seem flip, but we mean it |
| 100 | + sincerely: comments that include some tangible suggestion |
| 101 | + for improvement are tremendously helpful. |
| 102 | + |
| 103 | +<!-- links --> |
| 104 | + |
| 105 | +[sc21]: https://sc21.supercomputing.org/ |
| 106 | +[coordrepo]: https://github.com/hpc-carpentry/coordination/tree/main/conferences/SC21 |
| 107 | +[bofcodi]: https://codimd.carpentries.org/9-Y8OaVIT2qpb_P47TR7Lw |
| 108 | +[milestone]: https://github.com/hpc-carpentry/coordination/milestone/1 |
0 commit comments