Choose timezone
Your profile timezone:
→
Europe/Zurich
6/2-024 - BE Auditorium Meyrin (CERN)
Maria Girone
(CERN),
Tommaso Boccali
(INFN Sezione di Pisa)
Description
Meeting to discuss the current status and challenges of HPC integration in all regions.
There is a google doc for live notes
https://docs.google.com/document/d/1idgj-W9EX4KF6DTDihTLAwKcDtbYtA5-k0o54hx_pHA/edit?tab=t.0
Registration
Participants
1 - projects
a - CERN/WLCG provide a repository where we will ask national/sites/experiments/individuals to put material about
1 - their attempted / won / rejected allocations in the last 5 years (whenever possible / not under NDA)
2 - conference notes, about integration attempts
b - discuss and encourge at the WLCG MB level for the possibility for common projects (regular scale, development)
c - we still encourage bottom up R&D / exploration initiatives
d - explore development and benchmarking opportunities with support from HPC sites (for example Epicure)
2 - multi year access (using regular projects + automatic renewal) easier on the EU side than strategic
3 - community access (strategic access) - in EU:
a - long term allocations remain our ultimate goal; comm. access can be a way but not the only one
1 - it was explicitly a goal for IRI -- but not atm; putting some pressure there might help
b - use the FAs as the preferred pressure mechanism, then CERN can pick up (bring to the attention of the RRB)
b2- use the same mechanism to try and influence the future HPC machines
c - giving an AI goal helps a lot in increasing the probability
4 - IRI and EuroHPC FP - EFP - (granted to CSC)
a - we need a direct contact with these, and ways to interact (NOW!)
1 - understand what is being done now (existing apis, reference science domains)
a - Liz: no HEP names in IRI? Can we get someone? Paolo: CCE can ...
2 - try to enter the set of reference sciences and be able to communicate our needs and influence the process
3 - SCHEDULES
a - IRI has real use cases, so have to come to something usable < 2 y (not shorter because it's a volunteer effort)
b - EFP to be deployed on 9 sites in 1-1.5 y ... need to act soon
1 - Leonardo, Lumi, Meluxina, ... (most probably the peta-, preexa- and exa-scale sites)
2 - try to approach these centers where we know them
5 - ESPP document? white paper to be written anyhow and reused for ESPP?
a - older white paper (Maria et al ...) from 2020 + a draft of a white paper existing
6 - how to incluence future HPC centers?
a - CERN/WLCG provide a repository where we will ask national/sites/experiments/individuals to put material about
1 - their attempted / won / rejected allocations in the last 5 years (whenever possible / not under NDA)
2 - conference notes, about integration attempts
b - discuss and encourge at the WLCG MB level for the possibility for common projects (regular scale, development)
c - we still encourage bottom up R&D / exploration initiatives
d - explore development and benchmarking opportunities with support from HPC sites (for example Epicure)
2 - multi year access (using regular projects + automatic renewal) easier on the EU side than strategic
3 - community access (strategic access) - in EU:
a - long term allocations remain our ultimate goal; comm. access can be a way but not the only one
1 - it was explicitly a goal for IRI -- but not atm; putting some pressure there might help
b - use the FAs as the preferred pressure mechanism, then CERN can pick up (bring to the attention of the RRB)
b2- use the same mechanism to try and influence the future HPC machines
c - giving an AI goal helps a lot in increasing the probability
4 - IRI and EuroHPC FP - EFP - (granted to CSC)
a - we need a direct contact with these, and ways to interact (NOW!)
1 - understand what is being done now (existing apis, reference science domains)
a - Liz: no HEP names in IRI? Can we get someone? Paolo: CCE can ...
2 - try to enter the set of reference sciences and be able to communicate our needs and influence the process
3 - SCHEDULES
a - IRI has real use cases, so have to come to something usable < 2 y (not shorter because it's a volunteer effort)
b - EFP to be deployed on 9 sites in 1-1.5 y ... need to act soon
1 - Leonardo, Lumi, Meluxina, ... (most probably the peta-, preexa- and exa-scale sites)
2 - try to approach these centers where we know them
5 - ESPP document? white paper to be written anyhow and reused for ESPP?
a - older white paper (Maria et al ...) from 2020 + a draft of a white paper existing
6 - how to incluence future HPC centers?
There are minutes attached to this event.
Show them.
-
- 13:00 → 13:15
-
13:15
→
13:30
Experiments computing models, software and needs¶
-
13:15
Speaker: Dr Tommaso Boccali (INFN Sezione di Pisa)
-
13:15
-
13:30
→
14:10
Experiments HPC usage and perspectives¶
- 13:30
-
13:40
Speaker: Daniele Spiga (Universita e INFN, Perugia (IT))
- 13:50
-
14:00
Speakers: Alexandre Franck Boyer (CERN), Concezio Bozzi (INFN Ferrara)
- 14:10 → 14:20
-
14:20
→
15:00
Analysis and Discussion of the US and EU meetings¶
A common presentation by representatives from both meetings:
- Common Challenges
- Geographically specific issues
- What is missing
- Consensus on the list of common topics-
14:20
Speaker: Andrej Filipcic (Jozef Stefan Institute (SI))
-
14:40
Speaker: Paolo Calafiura (Lawrence Berkeley National Lab. (US))
-
14:20
- 15:00 → 15:10
-
15:10
→
15:30
Break 20m
-
15:30
→
17:00
Update on HPC common initiatives¶
-
15:30
Speaker: Deborah BARD
-
15:45
Speaker: Evangelos Floros (EuroHPC JU)
-
16:00
Speaker: Henrik Nortamo (CSC)
-
16:15
Speaker: Maria Paola Lombardo
-
16:30
Speaker: Concezio Bozzi (INFN Ferrara)
-
16:45
Speaker: Junichi Tanaka (University of Tokyo (JP))
-
15:30
-
17:00
→
17:40
The HPC landscape: exascale, post-exascale and AI: - first session¶
-
17:00
Speakers: Fredrik Robertsén (CSC), Pekka Manninen (CSC)
-
17:10
Speaker: Taylor Childers
-
17:25
Speaker: Dan Stanzione (UTexas)
-
17:00
-
17:40
→
18:10
Report from integration efforts at WLCG centers (USA time-zone)¶
-
17:50
Speaker: Alexei Klimentov (Brookhaven National Laboratory (US))
-
17:50
-
18:30
→
20:00
Network Drink 1h 30m
-
-
08:55
→
10:20
The HPC landscape: exascale, post-exascale and AI: - second session¶
-
08:55
Speaker: Dirk Hufnagel (Fermi National Accelerator Lab. (US))
-
09:05
Speaker: Guillaume Lechantre (GENCI)
-
09:15
Speaker: Florian Berberich (FZJ)
-
09:25
Speaker: Dr Riccardo Di Maria (CERN)
-
09:35
Speaker: Andrej Filipcic (Jozef Stefan Institute (SI))
-
09:45
Speaker: Jorik van Kemenade (SURF)
-
08:55
-
10:20
→
10:40
Break and group photo 20m
-
10:40
→
11:50
Report from integration efforts at WLCG centers (USA time-zone): (for the EU + Asia timezone -- 4 10 min talks)¶
-
10:40
Speaker: Hyeonjin Yu (KISTI)
-
10:50
Speaker: Ben Jones (CERN)
-
11:00
Speakers: Matthias Jochen Schnepf, Matthias Schnepf
- 11:10
-
11:20
Speaker: Daniele Cesini (Universita e INFN, Bologna (IT))
-
10:40
- 11:50 → 12:25
-
12:30
→
13:30
Lunch 1h
-
13:30
→
14:30
Path towards HPC integration¶
- How do we move beyond individual site connections?
- Timeline and schedule assessment from convergence discussion
- What do we need to maintain persistent long term collaboration?
- How do we establish a common R&D programs and keep them coherent?
-
13:30
Speakers: Gonzalo Merino (IFAE - Institute for High Energy Physics), Dr Ian Fisk (Simons Foundation)
-
14:30
→
15:00
Break 30m
-
15:00
→
16:00
Next steps (All)¶
- Document sections and outline
- Critical path items
- Contribution to ESPP
- Future meetings
-
15:00
-
08:55
→
10:20
