PANDA Grid Activities
D. Protopopescu
1, A. Amoroso
2, P. B¨uhler
3, R. Dosdall
4, P. Gianotti
5, C. Grigoras¸
6, J. Messchendorp
7, V. Mitsyn
8, D. Pantea
9, F. Ronchetti
5, P. Saiz
6, K. Schwarz
10, M. Al-Turany
10,
F. ¨ Uhlig
10, and J. van der Weele
71University of Glasgow, UK; 2INFN Torino, Italy; 3SMI Vienna, Austria; 4Forschungszentrum J¨ulich, Germany;
5INFN Frascati, Italy; 6CERN, Geneva, Switzerland; 7KVI Gr¨oningen, The Netherlands; 8JINR Dubna, Russia;
9IFIN-HH Bucharest, Romania; 10GSI, Darmstadt, Germany
Abstract
PANDA Grid is the present computing infrastructure of the PANDA@FAIR experiment, but concerted efforts are put into evolving it beyond that, into a complete and trans- parent solution for physics simulation, reconstruction and analysis, right at the fingertips of the physicist.
PANDA Grid
Experiments with very high data acquisition rates like PANDA [1] could not be thought to stand a chance without a Grid model for their computing needs. PANDA collab- oration established PANDA Grid as its computing infras- tructure [2].
PANDA Grid uses as middleware AliEn [3], developed by ALICE@CERN and used outside CERN now, besides PANDA, by other FAIR experiments, like CBM. The soft- ware is centered on PandaROOT [4], developed partly within the PANDA collaboration, partly within the larger FAIR community. Grid monitoring and data production su- pervision are done via MonALISA [5].
PANDA Grid encompasses now 19 sites, from 12 insti- tutes in 8 countries. The last three sites were added during 2008. While being an up and running system, PANDA Grid is continually growing, with 3 new institutes having already announced their intention to join in 2009.
A Tool for Physics
During the last year, through concerted efforts from a heterogeneous team of physicists and software developers, we brought PANDA Grid beyond mere ’computing infras- tructure’. Our aim is to transform it into a ’tool for physics’, that is a complete and transparent solution for physics simulation, reconstruction and analysis, with minimal overhead for the user (i.e. the physicist).
All the software for data simulation, reconstruction and analysis in our collaboration is based on PandaROOT.
In our computing model, the ’user’ software is based on ROOT macros using modular precompiled libraries from PandaROOT. The PANDA Grid infrastructure is used for software distribution via the package manager service (PackMan) provided by the middleware layer. We set up in 2008 a completely automatic system for package instal- lation, compilation and testing, supervised directly by the PandaROOT developers, in such a way that the user has to
customize only a couple of scripts before pursuing his/her own physics.
AliEn is the perfect tool for the job, as it already pro- vides advanced data production tools at the middleware level (various mechanisms for job split and merging, cat- alogue tags and triggers, file collections have been imple- mented [3]). But the addition of the MonALISA integration layer allowed us to develop in 2008 a range of supervision tools and command mechanisms that made PANDA Grid easy to administer and easy to use for large productions.
Close contact between the subgroups participating in this effort is maintained by numerous ad-hoc meetings as well as via dedicated workshops [7], hosted in turn by various participant institutes. Two PANDA Grid workshops were hosted in 2008, in Rome, Italy and Sinaia, Romania.
In 2008 we also had two data challenges [6], during which large quantities of data were produced via a full chain of simulation, digitization and reconstruction. This data will be analysed and used to improve the design of the PANDA detector. The GSI and JINR sites are at the time being the largest contributors in terms of computing power.
Since this is a crucial ingredient for creating the com- plete solution we aim for, during 2008, significant effort has been dedicated to perfecting the organizational frame- work for administration and usage of PANDA Grid. A production management scheme has been set in place (and tested during the data challenges), complete with user, data storage and scheduling schemes. An official participa- tion document - a so-called statement of service (SoS) - has been drafted to formalize the participation process in PANDA Grid. There is now a mutually beneficial col- laboration in terms of software development and Grid de- ployment between ALICE and the FAIR experiments using AliEn, with several joint meetings in 2009 [8].
References
[1] http://www.gsi.de/panda
[2] http://panda-wiki.gsi.de/cgi-bin/view/Computing/PandaGrid [3] http://alien.cern.ch
[4] http://panda-wiki.gsi.de/cgi-bin/view/Computing/PandaRoot [5] http://monalisa.cern.ch
[6] see ’Data Challenges’ in [2]
[7] http://nuclear.gla.ac.uk/grid-workshop/
[8] M. Al-Turany et. al, Grid Activities at GSI, 2008