CERN Accelerating science

On Friday 1 March, the EU-funded DataGrid project successfully passed the first year review performed by external experts appointed by the European Union.

In late 1990, a CERN computer scientist invented the World Wide Web to facilitate exchange of information between scientists working on different computers, perhaps at different sites. At the EU review, CERN, along with DataGrid project partners INFN/CNAF (Bologna, Italy), CNRS/IN2P3 (Lyon, France), PPARC/RAL (UK), and FOM/NIKHEF (Amsterdam/NL), demonstrated the first DataGrid test bed. The WWW is mainly aimed at the exchange of information while the Grid is concerned with the exchange of computer power, data-storage, and accessing large databases, without forcing users to search for these resources. Once "connected" to the Grid, the end user will see it essentially as one large computer system. For these reasons, many believe the Grid to be the most practical solution to the so-called data intensive science problem that must be overcome if the computing needs of scientific communities such as processing of physics data from LHC experiments are to be satisfied.

In about five years, LHC will start to produce an enormous stream of data that must be stored and processed. The "data current" is estimated to be a few petabytes per year, and in excess of 50,000 high-end workstations will be needed to keep all the physicists' analysis tasks going. Because no single HEP computer centre is able to provide for both the storage and computing facilities for the entire LHC operation, distribution of computation and data is required. The DataGrid project has begun to solve this problem by developing methods and software that will efficiently and smoothly distribute requests to use computing and storage resources offered by all computers centres connected to the Grid.

The EU review was a day-long affair at CERN. The 16 other member institutions of the DataGrid project were represented by on the order of 60 scientists in attendance. These project partners, together with the six main partners, are responsible for most of the technical developments including innovative middleware software, testbed deployment, and specification of application test use cases. The review culminated in a live demonstration in which jobs were submitted to the DataGrid testbed. A total of about fifteen jobs were submitted, mostly by representatives of physics experiments, but also by representatives of the Earth Observation and Computational Biology communities, who are also participating in the DataGrid project. The jobs were successfully distributed across the five computer centres that participated.

The EU reviewers congratulated the project for the excellent work done at both technical and managerial level, exceeding the expectations for the first year in some areas. They even complemented the few "disturbances" in some of the Grid services that occurred during the demonstration, noting that they would have been very suspicious if everything had worked flawlessly at such an early stage in the project.

The jobs submitted during the demo consumed in total less than one CPU-hour. The demonstration showed the capability of the Grid to distribute computational tasks, but came nowhere close to demonstrating how much computer power it can provide. Now that the demonstration is over, more users have been admitted to the Grid and experiments are beginning to conduct more demanding tests. This power will increase rapidly as the existing centres expand and new sites join the Grid.

Visit http://eu-datagrid.web.cern.ch/eu-datagrid/ for more information on the project.