Résumé / Abstract Journal-club_Univers

Séminaire Univers /
Seminar Universe

« Running a Planck analysis on Cosmology@Home: is volunteer computing really worth it? »

Marius Millea
Institut d'Astrophysique de Paris (Paris, France)

This talk will be extremely informal, in fact, half of it will be a hands-on demo. Its also a "pitch" to you, as I'm actively searching for anyone interested in running their science code on Cosmology@Home (C@H) and helping develop C@H further. C@H is a volunteer computer cluster; people interested in science from all of over the world let us use part of the spare CPU time (like their home computer when the screensaver is on) to run our calculations. In this talk, I will tell you how we are currently using C@H to test the LCDM model by asking whether, when analyzing different multipole ranges of Planck data (or equivalently, WMAP vs. Planck), the size of parameter shifts we see are consistent with expectations. This involves running analyses on thousands of sets of simulated data, each job sent out to one of the ~10k active volunteers at C@H. One might rightly wonder if using volunteer resources is actually useful? Is the computing time significant? Is it easy to use? Or is this just good outreach? I'm going to argue that thanks to a lot of work over the last half-year from me and a few BOINC developers (that's the software C@H runs on), pretty much anyone can easily use these resources to run their science. I'll give a hands-on demo of what it looks like to run your code at C@H (which will involve Docker, a cool new software I think every scientist deserves to know about). And I'll comment on the types of analyses that fit well within the constraints of C@H.
mardi 1 mars 2016 - 11:15
Salle 281
Institut d'Astrophysique de Paris
Pages web du séminaire / Seminar's webpage