Speaker
Description of the work
The production of air shower libraries is done in a organized way by a single
central team. Tools developped by the team members cover several steps needed
for succesful operations: preparation of input files for simulated showers according
parameters given by the collaboration, bulk job submission, output registration
in the Auger dashboard, check of results and job resubmission in case of failures.
Tens of thousands jobs were executed in the grid infrastructure. We discuss
main obstacles in job throughput increase and how we overcome them. Tools
must be often adjusted to reflect changes in the middleware.
We also present organization of simulated data preservation and how we
export them to users without access to grid Storage Elements. Custom
product SimDB is used as the database for SRB installed in CC IN2P3 Lyon.
Impact
More than 250 members of the Pierre Auger Collaboration can use libraries
of air showers simulated with different models for several energy and theta
angle values. About 40 members can process data files directly on the grid,
the rest can access files at the CC IN2P3 in Lyon, where secondary copies
of all successfully produced libraries are stored.
Overview
The astroparticle physics project Pierre Auger Observatory observes
Ultra-high energy cosmic ray air showers using fluorescence and surface
detectors installed on a large area in Argentina. The CPU demanding job
of Monte Carlo simulations is done using grid resources.
Conclusions
We describe our practices for running Monte Carlo simulations jobs on the
grid and distributed data management. Current achievements and issues are
discussed.
