Choisissez le fuseau horaire
Le fuseau horaire de votre profil:
GT09 Town Hall Meeting: Calcul, Algorithmes et Données
→
Europe/Paris
Maison des Sciences de l'Homme, Clermont-Ferrand
Maison des Sciences de l'Homme, Clermont-Ferrand
4 Rue Ledru
63000 Clermont-Ferrand
Description
Cette assembl
ée publique fait partie des activités de prospective du CNRS / IN2P3, visant à définir les priorités scientifiques des prochaines années. Après une première analyse des priorités perçues par un grand nombre de collègues participants et la formulation des principales questions concernant la feuille de route, nous souhaitons maintenant examiner ces priorités pour calcul, algorithmes et données.
La réunion a eu lieu dans le centre-ville de Clermont-Ferrand, à distance de marche de la gare.
Vous n'avez pas pu assister à la réunion? Voir les vidéos de la réunion !
-
-
1
Introduction
- Description of the aim of the prospectives IN2P3
- Goals of the Town Hall Meeting
- Explanation of next steps
Orateur: Volker Beckmann (CNRS / IN2P3) -
Computing Models
-
2
Sustainable computing model(s) for IN2P3
Tier-model, centralised architecture or data lake approach: which model or mix of models best answers to our scientific requirements?
Orateur: Ghita Rahal (IN2P3/CNRS) -
3
Data Organisation Management and Access (DOMA)
First results of DOMA (Data Organisation Management and Access), data lake with xCache, possibility of diskless computing centres
Orateur: Eric Fede (CC-IN2P3/CNRS) -
4
Heterogeneous infrastructures
The challenge of integrating HPC, accelerators (e.g. GPU) and commercial cloud solutions — is it worth the effort, what should be the focus, what are potential benefits?
Orateur: Luisa Arrabito (LUPM) -
5
Table ronde: Computing model evolution at IN2P3
Participants: Luisa Arrabito (LUPM), Sabine Crépé-Renaudin (LPSC), Dominique Fouchez (CPPM), Fabrice Jammes (LPC), Pierre-Etienne Macchi (CC-IN2P3)
Moderated discussion of the above topics and questions.
- What is the impact of experiment’s centralized or distributed computing models at the French level?
- What about heterogeneity/complexity of supported solutions/experiments?
- Is IN2P3 willing/has the capability to participate to the current R&D works?Orateur: Nadine NEYROUD (LAPP)
-
2
-
15:05
Coffee Break
-
Artificial Intelligence
-
6
Development of ML for Particle Physics. Training and CS collaboration.Orateur: Julien Donini (UBP/LPC/IN2P3)
-
7
Development of ML for Astroparticle CosmologyOrateur: Dr Emille Ishida (LPC-UCA)
- 8
-
9
Table ronde: What ML development is the most appropriate for an IN2P3-wide approach, beyond exchange of experience ?
Does it make sense to push for advancement in Machine Learning on the institute level, or are problems/solutions too specific for each project and lab? What does this mean for the effort on training and recruitement?
Orateur: Yann Coadou (CPPM, Aix-Marseille Université, CNRS/IN2P3)
-
6
-
10
Internet of Things — a key topic for IN2P3?
Distributed detectors and control systems are a reality in IN2P3 experiments (accelerators, telescope arrays, KM3NeT,...). What are the possible benefits in a coordinated IoT effort, e.g. in combination with Machine Learning, and shall this be a key element of the perspectives?
Orateur: David SARRAMIA (LPC Clermont-Ferrand)
-
1
-
-
Overall infrastructure needs / development
-
11
Virtualisation and academic clouds
Virtualisation is not only interesting for small and medium size projects, but also in general a way to preserve and share workflows. Benefits and drawback of academic clouds, such as the ones provided through France Grilles.
Orateur: Vincent Breton (CNRS-IN2P3) -
12
Table ronde: How to better coordinate between the different domains?
Participants: Vincent Breton (LPC), Eric Chassande-Mottin (APC)
Solutions for WLCG and large projects are not necessarily applicable for a wide range of smaller projects within IN2P3. How can we better satisfy the different needs, how should this be coordinated and what should be the focus?
Orateur: Maude LE JEUNE (APC)
-
11
-
Accelerators
-
13
Portability of code
What can be the way forward in using accelerators (GPU, FPGA) and HPC in our field? Is dedicated re-coding necessary, or are abstraction layer libraries like alpaka, kokkos, SYCL the way forward?
Orateur: David Chamont (LAL - IN2P3 - CNRS) -
14
FPGA (Field-Programmable Gate Array)
Challenges and benefits in the usage of FPGAs
Orateur: Jean-Baptiste Sauvan (LLR)
-
13
-
10:30
Coffee Break
-
Emerging Technologies
-
15
Quantum Computing
Status and possible impact of Quantum Computing for our sciences
Orateur: Bogdan Vulpescu (Laboratoire de Physique de Clermont) -
16
Table ronde: Emerging technologies — shall we lead or shall we follow?
Participants: Catherine Biscarat (L2IT), Gilles Grasseau (LLR), Bogdan Vulpescu (LPC)
Emerging technologies, like Quantum Computing, provide a large potential but also a great risk. Shall IN2P3 engage actively in these domains as contributor / developer, or rather observe the field and enter as a user in a later stage?
Orateur: M. Rodolphe Clédassou (IN2P3)
-
15
-
Additional aspects
-
17
Open Data: motivation, challenges and solutions
The pressure on the scientific community, to provide open access to data derived in their experiments, is rising (e.g. loi numerique, plan national science ouverte). How do we respond to that, what changes are necessary?
Orateur: Eric Chassande-Mottin (CNRS AstroParticule et Cosmologie) -
18
Software development and quality
WLCG and space projects impose strict quality assurance for developers. What are key points here? What could be the benefit for other, smaller projects to impose standards, rules, etc. Should IN2P3 provide centralised solutions and services in this context?
Orateur: Antoine Pérus (Cnrs/In2p3/Lal)
-
17
-
Final Discussions
-
19
Table ronde: Future effort: focused or wide open?
Participants: L. Arrabito (LUPM), F. Derue (LPNHE), E. Ishida (LPC), A. Moller (LPC), P. Verdier (IN2P3)
Throughout the topics (computing model, ML, emerging technologies) a main question is whether we should focus our activities on a few, central projects and topics (WLCG, large astroparticle projects) and find solutions for those, or whether a broader approach should be taken. This is also connected to questions like whether we concentrate our effort to build up and maintain expertise on the lab level, or whether we push stronger for IN2P3-wide solutions, e.g. through stronger reseaux and participation in IN2P3-calcul&données master projects? Is creating expertise-clusters on specific topics in certain labs a way to go forward?
Orateur: Laurent Duflot (LAL) -
20
Summary & Outlook
Short summary of the main topics raised during the Town Hall Meeting and outlook on the next steps to arrive at the final "Prospective Calcul & Données".
Orateur: Volker Beckmann (CNRS / IN2P3)
-
19
-
