Choose timezone
Your profile timezone:
EP R&D Software Working Group Meeting
→
Europe/Zurich
32/1-A24 (CERN)
EP R&D Software Working Group Meeting
Zoom Meeting ID
62893115044
Host
Graeme A Stewart
Alternative host
Andre Sailer
Useful links
Join via phone
Zoom URL
Software R&D Working Meeting Minutes
2023-05-17
Apologies: Andi, Paul
Room: Graeme, Erica, André, Leonhard, Swathi, Alberto, Juan, Vincenzo, Javier
Remote: Witek, Felice, Dalila, Pere, Marco, Brieuc, Juraj, Marta, Benedikt
Introduction (Graeme)
- [ ] Annual Report 2022: Check the list of authors!
- Budget would be given to the R&D overall, then an allocation to WP7 that we need to adjust to.
Faster Analysis (Javier)
Cluster group allows one to load metadata incrementally (N.B. 100GB is much larger than the experiments use in practice today).
Slide 5, RNTuple speedup is in evts/s.
CMS use std::unique_ptr (experiment EDMs are interesting and non-obvious!).
When new columns are added, RNTuple gives zeros as a default on readback when the column doesn't exist - could it not be undefined, like in a DB?
- This does happen for std::variant, but not for simpler types like float, when it's 0.0.
- Used by the experiments when a "rich" event crops up, that has additional data that's needed.
- At least the behaviour is well defined, so the experiment can manage this, and there is a call that can be used to see if data really existed, or if it's the "default".
Will the late schema extensions require revisions to UpROOT and UnROOT?
- Yes, the schema is not yet finalised - but we are close.
- However, pre-existing files will still remain readable.
For S3 the protocol itself wants larger page sizes (http, higher latency than DAOS), not just because of the particular backend hardware used in this test.
AOB
Next Meeting 21 June, Turnkey (the Summer Solstice meeting!)
There are minutes attached to this event.
Show them.
