The HEP Software Foundation

Advancing high energy physics community software

HSF Weekly Meeting #125, 8 February, 2018

HSF Weekly Meeting #125, 8 February, 2018

Present/Contributors: Graeme Stewart, Giulio Eulisse, Eduardo Rodrigues, Mike Sokoloff, Mike Hildreth, Benedikt Hegner, Helge Meinhard, Daniel Elvira, David Crooks, Jim Pivarski, Torre Wenaus, Riccardo Maria Bianchi, Liz Sexton-Kennedy, Marco Clemencic, Eric Lancon, Paolo Calafiura, David Lange, Frunk Wuerthwein, Mohammad Al-Turany, Oli Gutsche, Eric Lançon, Alessandra Doria

News, general matters

  • Contacted by the organiser (promoter?) of the FOSS Backstage conference.
    • “The conference on everything related to governance, collaboration, legal and economics within the scope of FOSS”.
    • Not immediately clear if this would be actually interesting for us, although the related Berlin Buzzwords event does seem well established (open source big data conference).
    • Apache Foundation are an event partner (does this make it more real?).
    • Call for papers until 18th February.
    • Graeme will follow up.

HSF/WLCG Workshop

  • General points

    • After discussion we swapped Security and Performance Costs WG sessions on Wednesday, adding an extra Performance and Costs hands-on session as well.
      • Barring any crises, agenda is final.
    • We will be able to have a contribution from SKA on Thursday morning, hopefully in person. This will eat into the closing plenary a bit, but we feel it’s a valuable contribution to include.
    • 87 registered - this is ok, not great.
      • Can convenors and HSF advocates invite people to come? Individual invitations take time, but have a much higher success rate than blanket emails.
    • Early registration closes on 16th February.
    • Mike S - could use some S2I2 conceptualisation funds to help people from the US to come for HSF pieces: younger participants or session convenors; labs included.
    • Sponsor contributions to the workshop.
      • Rittal sent a title and a short abstract:
        • STATE-OF-THE-ART PHYSICAL INFRASTRUCTURE FOR HPC: Enabling HPC through a modular physical infrastructure is a must to ensure the best performance and an effective TCO. Minimized operative costs, and a scalable investment are key success factors in implementing HPC efficient platforms.
      • Still waiting for the abstract from Dell/Intel, they said we will receive it tomorrow.
      • We don’t know yet the names of the presenters.
      • Will the sponsor presentations be included in the Technology Watch session? A. Yes.
  • Monday

    • Opening Plenary
      • In good shape - all talks planned.
    • Technology Watch
      • Planned the session now (for 90 minutes, 30 goes to vendors).
    • HEP Use Cases
      • List of topics and speakers is created - will contact people in the next week.
  • Tuesday

    • Data Management/Lakes
      • Session is now planned in Indico.
    • Frameworks and Infrastructure
      • Planning on three 30’ talks: broadly speaking dedicated to online aspects (including GPU/coprocessor management), offline aspects (including thread/process/node management), and HPC parallelization, trying to cover multiple experiments/frameworks.
      • Boundaries with WMS and Fwk Technical to be discussed.
    • Training (Parallel)
      • No news.
    • DPHEP (Parallel)
      • Mike H and Jamie S did some planning on the session: Certification, CWP goals, brainstorming.
    • Workload Management (Parallel)
    • Analysis Facilities and Use Cases
      • Agenda finalised and skeleton up online.
      • Potential speakers all contacted, 2/5 confirmed.
  • Wednesday

    • Programming for Concurrency and Co-Processors
      • Still seeking co-convenor.
      • Areas:
        • Trigger
        • GeantV
        • Parallel tracking
        • Parallel processing models - developer of Dask from anaconda (outwith HEP)
          • We discussed this proposal and it was well received.
    • Visualisation (Parallel)
      • Riccardo & Tom started to gather proposals and filling the agenda.
      • A potential talk from an outside-of-HEP institute: University of Ljubljana Computing Science department.
    • Simulation
      • Two sessions: faster simulation and improving physics.
      • Have a list of speakers and contributions prepared.
        • Some speakers will be remote, need a good vidyo setup.
      • Can put these into the agenda now - Andrea or Daniel will do it.
    • Security (Parallel)
      • Second convenor from CMS? Brian Bockelman still might help out.
      • Topics, for further drafting and sorting into talks:
      • Trust and Policies
        • GDPR
      • Authentication and Authorisation
        • Authorisation WG status/summary
      • Operational Security
        • Requirements for operational security in the future
        • Threat Intelligence
        • Discussion
        • Operational security technology requirements (SOC WG status & summary)
        • Traceability
        • Containers/Singularity
        • IPv6
    • Performance and Cost Modeling (Parallel)
      • Progress in planning in the WG meetings.
    • Technical Discussion on Frameworks (Parallel)
      • Need to clarify which technical aspects of frameworks should be discussed.
      • In the plenary want to tackle the boundary between nodes and processing framework.
      • Technical session would try to decompose where we are on frameworks and address the CWP roadmap goals.
        • E.g., is TBB still an optimal choice?
        • What is a good cadence for updating framework components?
    • Software Development (Parallel)
      • Martin still in Japan. Giulio starting to think about possible discussion topics:
        • Project Incubation:
          • Does the Incubator model of Apache make sense for HSF?
          • As a project leaders what would make me move to HSF?
          • As a HSF reviewers what are the minimum requirement for incubating a project?
          • Graeme: can HSF provide a source of technical knowledge (design, implementation).
        • Project discussions:
          • Packaging requirements, what is missing? How do we scale to multiple project which have completely different timescales and requirements in terms of software stacks? How do we integrate into mainstream?
          • Static analysers. All the LHC experiments / projects use some form of static analysers, mostly based on clang. How do we contribute back all those efforts so that everyone can profit from that?
          • Profilers. How do we share and compare profile data between different tools and experiments?
          • GVFS: git has become the defacto standard VCS in HEP and in the world. Is GVFS the next big revolution in the way we work? If yes, what are the implications / possibilities?
        • Becoming mainstream:
          • What are the OpenSource conferences / workshops we should try to have a more “active” presence?
            • Python ones?
            • FOSDEM?
          • Geek watch: how do we share information about the latests tools and practices? Mailing list?
  • Thursday
    • Closing Plenary
      • Anticipate SKA contribution here: 20-30 minutes.

CWP

  • General Matters and Roadmap

    • arXiv update a little delayed - announcement was late and waiting for CMS resource plots with the correct axes labels (MHS06s -> THS06s)
      • We now have all the pieces though, so this can be done ~now.
      • Added 10 new signatures since last week.
    • Computing and Software for Big Science - discussed with Markus Elsing. He did not think it would be a problem to update the signatures after submission, but is confirming with other editors and Springer.
      • The lead time for publication is >8 weeks, so we would be ok to take signatures until the Naples workshop.
  • Publication strategy for Individual WG Papers

    • Simulation

      • There is a final version of the long note; circulating amongst the authors and calling for supporters.

PyHEP Workshop

  • Good progress with agenda and finding speakers.
  • Sponsorship from PSF awaiting news (expect anytime within the next 2 weeks).

Activity updates

Licensing

  • No news.

Training

  • Pete Elmer is putting together an NSF proposal that might find some HEP community training under the HSF umbrella.
    • Will be useful to discuss this further in Naples.

Packaging

GSoC

  • Proposals in, waiting for Google to confirm CERN-HSF as an organisation in the GSoC 2018 programme, which should happen by Feb. 12th.