URL: https://www.desy.de/aktuelles/news/archiv_vor_2010/2004/teilphys_1211/index_ger.html
Breadcrumb Navigation
HERA Experiment ZEUS Achieves Successful Mass Production on a Computing Grid
R. Mankel (DESY): The computing load was shared between the sites of DESY, Rutherford Appleton Laboratory (RAL, Oxfordshire), ScotGrid (Glasgow), University College London (UCL) and University of Hamburg. The production demonstrates feasibility and the huge potential of grid technology for an experiment that is currently taking data.
ZEUS is one of the three active experiments at the electron-proton storage ring HERA at DESY in Hamburg. Computer simulation of electron-proton (ep) collisions is essential for the extraction of physics results from the real data. Many years ago, the ZEUS collaboration had already been pioneering distributed event simulation with a homemade tool named “Funnel”. This mechanism has allowed producing most of the simulation outside of DESY at collaborating laboratories during the HERA-I era.
The recent HERA luminosity upgrade results in an impressive increase of interaction rates, resulting in a largely improved physics reach. In consequence, also larger samples of simulated events are required to allow full exploitation of the new data. Furthermore, due to the addition of new tracking components in ZEUS, notably the micro-vertex detector (MVD) and the straw-tube tracker (STT), the new detector setup is much more demanding in terms of computing power. In the worst case, a deficit of simulation capacity could lead to significant uncertainties in the analysis of the data. It has therefore become important to more than double the existing resources for simulation. Fortunately, the advent of “grid technology” promises a very attractive way to meet this challenge.
The term “grid” has been coined by Carl Kesselman and Ian Foster in the mid-1990s describing a vision of a networking infrastructure which provides ubiquitous access to large computing resources through a network, in direct analogy with the well-known electrical power grid. While the world-wide-web (WWW) is only designed for retrieval of comparatively small amounts of information, a “computing grid” should provide access to massive computing power. During recent years, the High Energy Physics community has spent considerable effort on development of software frameworks that allow the construction of grids. A very recent development is the software engine of the “LHC Computing Grid (LCG)”, which is already in use at about 80 connected grid sites.
Thus grid computing offers entirely new perspectives of resource sharing on a world-wide scale. For this reason, during the last year the ZEUS simulation software suite has been adapted to work also in a grid environment. This work has been performed in close co-operation with the DESY IT group, who has set up a core system that is now serving as the “hub” of the ZEUS grid. Besides DESY, several other international grid sites are at present actively running ZEUS production, including computer centres of RAL (Oxfordshire), ScotGrid (Glasgow), University College (London) as well as nodes of the University of Hamburg. Fig. 1 shows a snapshot display of the number of jobs running at a grid cluster at RAL for the registered virtual organisations. The display visualises the active jobs of ZEUS in yellow. Also job activities of the experiments ATLAS and LHCb, both of which are currently under construction at the CERN Large Hadron Collider (LHC), are displayed.
After commissioning of the ZEUS production tools in the grid environment, the production mark of 1 million ep collision events has been surpassed recently, achieving an important milestone for the offline computing of the experiment.
The event datasets produced on the grid have passed the regular quality tests successfully. Fig. 2 shows an event display of a ep collision simulated on the grid. It is worthwhile to note that the samples ZEUS produces on the grid are not merely generated for testing purposes, but are actually going into immediately use in the ongoing physics analyses.