CERN Accelerating science

Abstract

The DataGrid project is moving towards the production of a stable resource that scientists all over Europe will be able to plug into and use.

When the European DataGrid (EDG) project started, almost two years ago, one of its main aims was to go beyond the R&D phase and demonstrate a "production quality" computing Grid. Such a Grid would give scientists unprecedented computing power to tackle major challenges, such as modelling climate change or analysing genomic data, using a distributed network of computers involving major computer centres all over Europe. Production quality means not just a proof of principle, but the deployment of a stable resource that European scientists would be able to plug into and use on a regular basis.

In the last two years, those ambitious goals have come a lot closer to reality.

Already in March this year, during the first EU Project review, a demonstration was made of the first release of DataGrid middleware - the software that makes a Grid of computers work together seamlessly. Jobs were submitted to the DataGrid Testbed1, showing the capability of the grid to distribute computational tasks to the most appropriate resources, by matching the set of requirements specified by the user to the characteristics and status of the available resources (CPU power, memory size, CPU load).

During the CERN school of Computing (15-28 Sept 02, Vico Equense (Italy)), 80 students have been exercising the functionality of the software, submitting jobs during the hands-on session dedicated to the Grid.

A new release of the Grid middleware was officially made on November 11th, with greatly improved support for large file transfers, better tracking of applications as they run on the Grid and a more stable information system. Several key features for production have been added, such as simplified access to mass storage systems, an easier software installation mechanism and user friendly job submission facilities.

The software is currently being installed on hundreds of computers that make up the EDG production testbed. Initially limited to five European sites, the testbed has recently been enlarged to approximately 20 sites across Europe, including Poland, the Czech Republic, Spain, Portugal, Germany and the Nordic countries.

In the months to come, the focus and resources of the project will be concentrated on supporting the LHC experiments that are preparing for the future data taking and analysis by running the so-called Data Challenges, a set of computing tasks concerning software simulation of physics processes. The status of this work will be reported at the second EU project review scheduled for the 4th-5th February 2003.

An ATLAS-EDG Task Force was put together in August 2002 with the aim to assess the usability of the EDG testbed for the immediate production tasks. The first phase of "Data Challenge 1" has been already performed using various Grid tools and results have been presented during the September 19 Task Force meeting. The second phase, starting in October 2002, will use Grid tools as much as possible.

CMS is starting now similar usability evaluation, and the same is happening in the other LHC experiments and will be supported by the DataGrid experts. The hope is of course to contribute middleware technology and testbed infrastructure to the overall LHC Computing Grid project.

In the mean time, Earth Observation scientists are busy developing the EO Grid infrastructure, and deploying a first Grid-based prototype for collaborative processing and validation of atmospheric ozone levels measured from ESA satellites. Biomedical applications, which were already successfully demonstrated at the time of the first project review, continue to be developed and enhanced.

A practical introduction to grids and to the EDG software is available for anybody who participates in the "DataGrid tutorial", a 'travelling' event that, after having been featured already twice at CERN in October, will move to several institutions in the project members states. The tutorials are organized over 2 full consecutive days and include both a theoretical introduction on general grid concepts, user interface commands, grid components, etc. and hands-on exercises covering user interface operation with job submission, aspects of data management and job monitoring. No prior knowledge is required for the participation.

For more information on the DataGRID project see:
http://www.eu-datagrid.org

For the programme and schedule of the tutorial:
http://eu-datagrid.web.cern.ch/eu-datagrid/Tutorial/tutorial.htm