OpenDreamKit final review (rehearsal day 2)

This pad is for taking notes during the second day of rehearsal for OpenDreamKit's Final review.

TODO

  • WP presenters: please put your slides in the web page sources for the review and add links from here to your WP slides

  • lightning demos: please add links to your notebook on nbviewer + github

  • demos: please edit opendreamkit.org/try/ to point to your repo

  • Organize taxis (Steve), starting to leave at 7:30, arriving 8:20

  • Printouts (6 copies; three for reviewers, one for PO, one for potential extra person, one for Nicolas)

    • Updated agenda ()
    • Glossary

About WP presentations

Suggestions from our reviewers:

We need to understand not only the achievements of the project as a whole, but also to know whether this kind of project is viable in the future. Here are three suggestions:

  • Stick to high level presentations on work packages, without going too much into the details
  • Focus the presentations more on problems encountered
  • explain what the future generation of project can learn from our experience

One of the reviewers further suggested:

The project was funded on the idea of creating a collaborative environment. Therefore, the success of the project should be measured against this objective. He therefore suggests to make a special focus on:

  • How to evaluate Jupyter Notebooks as system by which one can collaborate internally
  • Same question for external collaboration

Mockup Agenda

Today, we will try to follow the agenda of the review day, starting from the second item.

  • 9:00: welcome by the PO
  • 9:05: around the table
  • 9:10: What's new on opendreamkit.org Nicolas M. Thiéry
  • 9:20: Story: Balthazar's use case Nicolas M. Thiéry (TODO: includes live collaboration)
  • 9:45: WP4
    • first part of talk (Min), roughly 20 min
    • lightning demo teasers + stories (integrated as pdf into Min's talk, 1-2 slides each), roughly 5-10 min speakers should be up with Min from the beginning so that the transitions are smooth
      • Demo teaser: 3D visualization in Jupyter (Marcin, Min), Micromagnetics (Maryian)
      • Demo teaser: Application of Jupyter to enterprise training (Olivier)
      • Demo teaser: A VRE using Jupyter and GAP (Alex): [slides]
      • MAYBE: Story on Publishing computational logbook (Luca)
    • second part of talk (Min): future of Jupyter + questions, roughly 10 min
  • 10:30: Coffee break
  • 10:50: WP5: 30 minutes (including 2-3 lightning demos) + questions
  • 11:30: parallel demos (everybody but reviewers and presenters leave the room)
    • Alex: GAP
    • Olivier: Logilab teaching
    • Marcin: 3D
    • Marian: Micromagnetics
    • Min: nbdime, collaborative notebooks, more?
    • Steve: GAP HPC
    • David: Singular HPC
  • 12:00: lunch
  • 12:45: WP1: (Izabela + Nicolas), only site leaders are needed
  • 13:30: WP3: 20 min talk + questions (Luca)
  • 14:00: WP6: 20 min talk + frontal demo + questions (Michael)
  • 14:40: Coffee break
  • 15:00: WP2: 20 min talk + questions + slack (Viviane + Erik)
  • 15:30: Wrapup (Nicolas) + discussion with reviewers
  • 16:00: Group Picture

Every work package leader

  • coordinates which people are on stage to during the presentation (e.g., to take questions)
  • at the beginning of their talk, asks everybody else who was involved to stand up

Last year's schedule: see https://opendreamkit.org/meetings/2018-10-28-Luxembourg/ProjectReview/

Talking Points for the Review

Let's coordinate a few simple, concise, quote-worthy messages that we can repeat in

  • communications with reviewers and PO
  • our presentations
  • the wrap-up discussion We should focus on messages that the reviewers and PO can remember and report up the chain.

Talking points:

  • Open Source Software must be recognized as a critical e-infrastructure category for the EU. (Contrary to others, Open Source Software is decentralized and less tangible and therefore often overlooked.)

  • Funding for Research Software Engineers is essential for scientific e-infrastructures. (Contrary to funding for PhD students and senior researchers, RSEs are often overlooked.)

  • OpenDreamKit-like VREs have the potential to massively transform how mathematical sciences are researched, taught, and applied.

  • FAIR research data management will play a huge role in the future - it is particularly challenging for the mathematical sciences. (Contrary to other sciences, math data is more complex and requires research software to manage.)

Notes about presentations

Include timing information, etc …

Talk: "What's new in ODK"

  • Start with more general considerations, introducing the entire meeting
  • before diving right into the website and use case videos
  • Tell the story about the comics

Story Baltazar (15min)

  • motivation for implementing the mod rep algorithm ? (besides beauty)
  • slide 3.4 capitalize paseshnick
  • Don't say/write "in a single system", use "in a single (research) environment" instead
  • too much maths in the intro
  • pics of Balthazar
  • too many "and there is more" -> try to compact it into fewer
  • add VRE, toolkit, ODK keywords everywhere
  • add colors esp in the first slides

WP4

  • Call for WP4 demoers

  • Make WP4 people stand up

  • slide 4: Don't read the whole text

  • "this project got the ACM award": project is ambiguous; use "The Jupyter project"

  • slide 7: practice what you want to say here

  • slide 8: add "…" to figure to indicate there are many kernels, not just 3

  • slide 10: in/with distinction is good but was not clearly explained - give a few examples for each

  • slide 13, 14, 15: mention that you have demos on these

  • maybe run or tease the demos inside the talk

  • slide 17: say more or skip entirely

  • maybe merge slide 18+19

  • slide 22: involved in jupyter.mathhub.info also @kohlhase @florian-rabe

  • slide about deployments: write KPI and give a number

  • after demos: mention/show a shiny VRE demo as a transition to discussion on future

  • slide 30: science and maths -> science, technology, engineering, maths (STEM)

  • Future: briefly describe the bossee proposal (show the WP picture + consortium), how everybody wanted to be part of it, why we thought we had a good story and were disapointed this did not fly; asking for advice;

VRE Gap perspective Demo

  • intro: Min's talk mentioned GAP as one of the kernels, that lets you transition easily
  • Add a screenshot of using GAP before: terminal + XGAP
  • Mention / show that all the pictures are computed by GAP

WP5 presentation

  • Capitalize HPMC

  • "mathematical computing" should perhaps include R and C

  • "all digits matter":

    • Most of the time, all digits matters
    • There is more than digits / many objects are not just digits

  • Everything reduce to linear algebra; this quote may bring confusion; there are other key building blocks (arithmetic, combinatorics, …) that were worked on during ODK.

  • Comment about "subcubic" being considered forever as impossible in the HPC community for huge clusters, and things changing?

  • Slide 4: "our toolkit is made of systems and components"

  • Slide 9 "arithmetic" spelling

  • pronunciation of "promising"

  • "16 papers" – actually published in refereed journals? Say so if so.

Steve's demo: Meataxe 64

  • mention that the parallel demo session will be right after your teasers
  • zoom in the browser to increase font
  • display the order of the Monster group in bold
  • Should Steve be presenting the GAP-HPC slides just before his demo?
  • provide screenshots of the demo for inclusion in Clément's PDF

Daniel's demo:

  • "back to the 80's way of doing things"
  • you can speak louder
  • use a lighter theme for Jupyter (white background for better readability)
  • the before-after contrast would be nicer if you ran Singular in a shell instead of browser

###Lessons Learnt:

  • Premature focus … This is not a general statement, but something about one of the actions (linear alge); specify which.

WP6 presentation

Start: 12:06

  • add slide numbers
  • generally feels a bit unrehearsed

History slides

  • MitM picture: put actual system names instead A, B, C, …
  • during summary mention the words "documents" and "computation" already, which then occur in the tetrapod
  • first tetrapod picture: Organization -> MitM, Tabulation -> Data
  • "I'm not gonna talk about that - ask John" sounds a bit dismissive Maybe: "John is the expert and will happily tell you more" ? (Agreed – JC)
  • Mention tetrapod reachout events at CICM?

Isabelle library:

  • specify task number
  • The used ressources are not that impressive; mention them? rephrase "used resources" -> "run time"
  • Impact?
  • L4 is unix but not linux

FAIR slide:

  • arrow before "Questions" in wrong place
  • mention "lazy data"?

Nutshell slide

  • "your data" -> "your dataset"

State of Play slide

  • "The DB researchers are very interested DB aspects" in the missing
  • It's not convincing that MathHub Data wouldn't host all small groups
  • mention more on Future
  • tease the demo here

Persistent Memoization slide

  • not completely different, both are about locaL/short-term resp. global/long-term storage of data

  • I found this slide quite confusing (JC)

  • "Python for us means Sage" -> "our main use case is Sage"

Lessons Learnt

  • "which my colleagues from WP5 hate" sounds bad
  • other difficulty: using MitM currently adds a large dependency (MMT, …)
  • "high road" sounds dismissive of our other VREs; maybe emphasize the envisioned advantages of the approach and why they are hard to accomplish can say that WP6 had a stronger focus on research/exploration, when other WP had a stronger focus on delivering production ready solutions

Conclusion

  • "two approaches/attempts" -> "two points in a spectrum of which methods to use" emphasize methods are complementare, not competing
  • It's a lot of work and it's a long road to production ready and wide adoptioon
  • make possible -> made possible

WP2

Start: 15:16-43 + 3 minutes break

Comics: larger picture

There is some duplication about comics and videos with Nicolas intro. Synchronize and decide.

quote on impostor syndrome: put on the slide who said that (It currently looks like you said it.) same with quotes on later slides

Speaking of Sage+GAP training in Nigeria, remember to mention Software Carpentry training preceding it

24 minutes total; about 20 minutes real time; great. With a bit of language polishing this can probably be brought down to 18-19 minutes for a bit of extra margin.

WP 3

Started 16:05

  • Generally speaking could be more synthetic on KPI's

  • Opening slide: a real researcher would never ask for a VRE as such!

  • 9.2 say how many/which components were packaged

  • 11.1: showcase 'part of' not 'all of ' the work of WP5 D3.11 is more related to WP5 contributions delivered at M36 (D5.12) and not really on those presented today. Emphasize the pb of composing parallel libs into a high level vre.

  • 13.? Fix: Code qualty

  • Future: what are the main challenges? E.g. on keeping packaging efforts; how better is the situation now? Incremental enough?

  • [x]13.2: 80000 -> 80k

  • 14.1 add "impossible without funding for RSEs"

    • move "This would have been" up so that it applies to all 3 items

  • Lessons learned: those are not really lessons we learned; we knew about it.

  • "We need to be constantly.."

Select a repo