SIMCODE-DS PI: Marco Baldi (marco.baldi5@unibo.it) Presenter: Matteo Nori (matteo.nori3@unibo.it) Physics and Astronomy Department Bologna University
Simulations for Precision Cosmology We are about to enter the epoch of “Precision Cosmology” (testing the cosmological model at ~1% precision) In the next decade, costly space missions such as the Euclid satellite will map the Universe with unprecedented accuracy Precision Cosmology NEEDS large computer simulations: cDE-CDM f(R) gravity fNL(k) mν ≠ 0 Last decade: Planck Next decade: Euclid ΛCDM
We need large storage facilities! In the next few years we will need to develop simulations with more than a trillion (i.e. 1012!!!) particles: Euclid Full Scale Runs One snapshot of a Euclid Full Scale simulation: ~180-200 TB The complete simulation: a few PB Data can have a variety of formats, including HDF5 and FITS Data should be used for scientific exploitation by a broad community and for an extended period of time (~5-10 yrs) PROBLEMS: - Where to preserve such big data for such a long time? - How to distribute/transfer/manage efficiently these data? - How to allow a full scientific exploitation of the simulations? NEED of a dedicated infrastructure with cutting-edge technologies!
The SIMCODE project SIMCODE is a computational cosmology project funded by the Italian government through the High-Qualification programme SIR with a 1/2 Million Euro grant for a period of 3 years. SIMCODE will perform large cosmological simulations and investigate their observational footprints. SIMCODE therefore represents an intermediate step towards the development of pipelines and tools that will be employed in a few years for the next generation of simulations. The present SIMCODE-DS (Data Storage) pilot aims at testing possible strategies and tools for data preservation and distribution in view of these long term needs of the computational cosmology community.
The SIMCODE work flow Data Sharing and distribution Level 0: raw data (within the collaboration) Level 1: processed data (external collaborators) Level 3: science-ready data (anyone) Data production Large HPC facilities Long-term data storage EUDAT facilities raw data: snapshots processed data: maps and catalogs Data Analysis & Benchmarking access raw data Additional data analysis data subset recover
Please visit www.marcobaldi.it/SIMCODE