Choose timezone
Your profile timezone:
System performance modelling WG meeting
→
Europe/Zurich
31/S-028 (CERN)
Participants
- Local: Andrea Sciabà, Dirk Duellmann, Markus Schulz, Helge Meinhard, Domenico Giordano, Catherine Biscarat
- Remote: Gareth Roy, Michele Michelotto, Oxana Smirnova, Concezio Bozzi, Renaud Vernet, David Lange, Andrea Sartirana, Jan Iven, Frank Würthwein, Pepe Flix, Tommaso Boccali
Resource estimation processes
Concezio explains how LHCb is estimating their resource needs for Run3. His summary has been attached to the agenda.
Markus asks what drives the balance between fully and partially paremeterised simulation; Concezio says that it is driven by physics, as full simulation is of course preferred to minimise systematics. Semi-inclusive analyses, on the other hand, require huge amounts of MC, which would be unsustainable to generate using full simulation. Very, very roughly, fast simulation might amount to 50-70% of the total (not to be quoted).
Then David shows the process used by CMS, summarises the largest tasks and the main input parameters.
It is expected that datasets used for analysis will be much more compact at HL-LHC, and analysis will use just a small fraction of the total resource needs.
The expected pileup should not exceed 200 (median) and it could easily flat out at this value, as the machine will be capable of easily doing more. All simulations are performed using a Poisson distribution for the number of pilup events, not just a fixed value.
Status report on tasks
We go through the listed tasks and discuss any progress made.
On the estimation of costs for a data centre, Gareth says that at his site they do not have to pay for electricity, but he is still interested in estimating such cost. Markus argues that there is no need to estimate a cost that does not exist as it is expected that different sites can have very different situations and it is useful to see these differences. Andrea Sc. argues that what really matters is to estimate electricity consumption. Pepe adds that he did this estimation for the Nordic Tier-1 and produced a spreadsheet that includes all costs (including network).
For the other tasks, see the notes in the task list document.
Next meeting is scheduled for January 10th.
There are minutes attached to this event.
Show them.
