Choose timezone
Your profile timezone:
EP R&D Software Working Group Meeting
→
Europe/Zurich
Zoomlandia
Zoomlandia
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
Introduction
- Additional conference, Connecting the Dots, 31 May - 3 June, in Princeton
- Abstracts by 19 March
- Next meetings we shall switch to a hybrid mode, in 32/1-A24, which is allowed a capacity of 20 people in-person
Efficient Analysis
-
Pere, bad performance of HDF5 is remarkable; surely because you had to adapt to the data model to HDF5? Could you use RNTuple with other scientific data?
- Javier, some people in FNAL are trying to experiment with this
- Graeme, in astronmy, data stored is very different - very large telescope image frames, large reads of complete “events”
- Jakob, much of the popularity from HPC integration (sharding, parallel access; well supported by HPC centres), hard to evaluate RNTuple in that scanario as we don’t have access to these machines
-
Graeme, slide 29, how is it ensured that the data is local?
- This is hard coded into these jobs
- Tests from DAOS (slide 34) show that jobs running with a high-speed backend can get really good read speeds without this artifical pre-placement
There are minutes attached to this event.
Show them.