CERN Accelerating science

ATLAS Slides
Report number	ATL-SOFT-SLIDE-2018-956
Title	Usage of Salomon HPC in the ATLAS Distributed Computing
Author(s)	Svatos, Michal (Academy of Sciences of the Czech Republic, Institute of Physics) ; Chudoba, Jiri (Academy of Sciences of the Czech Republic, Institute of Physics) ; Vokac, Petr (Czech Technical University in Prague)
Corporate author(s)	The ATLAS collaboration
Collaboration	ATLAS Collaboration
Submitted to	2nd Users' Conference of IT4Innovations, Ostrava, Czech Republic, 7 Nov 2018
Submitted by	[email protected] on 23 Oct 2018
Subject category	Particle Physics - Experiment
Accelerator/Facility, Experiment	CERN LHC ; ATLAS
Abstract	The ATLAS Experiment, along with other LHC experiments, requires huge computing capacity to achieve its physics goals. Its results were presented in more than 1000 papers. WLCG Tier centers connected via Grid (i.e. grid sites) provide majority of computing and storage capacity. Few years ago, it became obvious that standard grid computing resources are not sufficient if a flat budget is assumed. Therefore, new capacity needs to be found outside of of grid sites, e.g. in HPCs or clouds. The ATLAS Experiment uses above-quota computing resources of the Salomon HPC. The job submission is performed via ARC Compute Element (ARC-CE) installed at the grid site in Prague. The ARC-CE translates job requirements of ATLAS jobs into a script for a batch system which is then executed via ssh. Scripts and input files are shared between the grid site and the HPC cluster via sshfs. The ATLAS software is rsynced from the grid site to the HPC's scratch space once a day to ensure availability of recent software. This contribution explains settings of the current system and reports experience with its running.