Intensive Computing on the Internet:
DATA-GRID, A New European Computing Grid Project

Paris, March29, 2000

The new European computing grid project aims to provide users with access to intensive computing resources using a very high-speed network and large-scale pooling of resources, and to make this as easy as plugging an appliance into an electrical outlet. This project developed in response to the needs of the major particle physics experiments that the CNRS decided to support recently, alongside CERN. Thanks to this network, it will be possible to process much larger volumes of data (future particle physics experiments will require 10,000 times more volume). In addition, it will probably lead to other applications, in other sectors of research (genome studies, climatology, etc.) and offer vast possibilities in the industrial and commercial sectors.

A high-speed computation grid works like an electrical network, which provides all users with all the resources they need via an extremely simplified interface (an electrical socket) that is more or less standardized throughout the world. The full complexity of the underlying network (from the electrical power plant to the individual) is completely hidden. In addition, users can change the amount of electricity they consume abruptly with no prior warning (if they are electric company customers). With a computation grid, computing power and storage capacity are practically unlimited, since all the grid's resources can be mobilized if necessary. The grid makes it possible to effortlessly put a locally developed application into intensive production, and to better share available resources (in computing centers and laboratories, or among the various sites of a single company).

Two factors are indispensable for a grid to be feasible: very high-speed networks over long distances, in order to have a sufficiently powerful underlying network infrastructure, and the capacity to manage service quality in order to guarantee that the throughput required by applications will be available. The development of communications technology has made this possible today. This new possibility is likely to bring about substantial changes in computing approaches in three major areas: intensive computing, visualization, and large data bases.

The material resources—grid nodes (computing centers, laboratory computers, etc.) and network—already exist for the most part. What remains to be built is the complete software infrastructure on which the grid is based. This infrastructure covers, among other things, identification and security, software for resource arbitration, application monitoring, quality assurance, and user interface. Security and identification in particular are fundamental, highly complex issues which are also of key importance for the future of e-commerce.

The CNRS is already involved in the development and production of computing grids through IDRIS, its national computer resource center for all basic research. A European partnership (the EUROGRID project) between this resource center and other national centers (in particular, in Germany and England) proposes to set up computing grids across national boundaries. This initiative would benefit certain disciplines for which European cooperation is especially well-structured, and lead to closer collaboration between very large computing facilities in order to optimize efficacy.

CERN, the European laboratory for particle physics, decided to base the computing architecture it plans to develop for its future experiments with the new LHC collider on the grid concept. This project, called DATA-GRID (Research and technology development for an international Data-grid), is unique due to the very high volume of data—100,000 tera-bytes, i.e., 100 million billion bytes, compared to 10 tera-bytes today—which will gradually be implemented, and the computing power necessary for processing them.

The CNRS is doubly involved in DATA-GRID:
o IN2P3—the French National Institute for Nuclear and Particle Physics—in partnership with DAPNIA (Department of Astrophysics, Nuclear Physics, Particle Physics, and Instrumentation associated with the CEA, France's Atomic Energy Commission) is naturally involved in the project directed by CERN. It will contribute mainly by assessing the relevancy of the model for the specific needs of high-energy physics, but it will also provide platforms of large-scale tests for the entire project. The IN2P3 computing center in Lyon will thus be an important node in the future grid.
o The CNRS is going to expand the base of DATA-GRID applications by studying the impact of the grid concept in other disciplines. Several areas in the various departments that could potentially be interested by this new approach have been identified: processing of satellite data, bio-informatics, genome studies, climatology, medical imaging, documentation data bases, programming, and parallel computing.

The CNRS will moreover contribute its expertise in high-speed networks and international networks and its links with Renater, as well as in security issues and controlling access to resources. A partnership could be developed with INRIA (French National Institute for Research in Computer Science and Control). Regional synergy will be created in the Rhone-Alps region, in particular with the presence of the IN2P3 computing center in Lyon, the laboratory of UREC (CNRS Networks Research Unit) in Grenoble, and the fact that CERN is nearby, in Geneva.

DATA-GRID brings together approximately ten European countries, as well as a particle physics computing center in the US. Several large-scale computing grid projects are in preparation in the US. In Europe, the national projects being contemplated in Great Britain and Italy are also associated with the above-mentioned projects.

Following encouragement from the European Community, a request for substantial funding will soon be sent for the DATA-GRID and EUROGRID projects. Applications will also be filed in response to the invitation for bids from the ministry (RNRT, National Network for Research and Telecommunications) and the new RNTL (National Network for Software Technologies).

In addition to its utility for scientific disciplines, it is very likely that the computing grid can be used for industrial and commercial applications:
o Industry will be involved in the development of the software that is essential for the creation of the grid;
o Businesses that have significant computing needs and private networks will be able to use internal grids;
o It will be possible to provide computer resources on request (computing, storage, applications, etc.) to companies with substantial needs that do not wish to invest in computing centers or private storage.

Contact at IN2P3 / CNRS (Institut national de physique nucléaire et corpusculaire):
Geneviève EDELHEIT
Tel: +33 1 44 96 47 60
E-mail: genevieve.edelheit@cnrs-dir.fr

Contact at CNRS Press Office:
Stéphanie BIA
Tel: +33 1 44 96 43 09
E-mail: stephanie.bia@cnrs-dir.fr