Ten-155: the current European research network. Credit: TEN

The 15 member states of the European Union have approved an 80 million euro (US$73 million) upgrade of Europe's research Internet networks. The move will provide the infrastructure needed to begin work on the concept of an advanced research computing ‘grid’.

A bid for preparatory work on a research grid was submitted to the European Commission last week by the coordinator, the European Laboratory for Particle Physics (CERN), with five other national partners, including France's National Centre for Scientific Research (CNRS), Britain's Particle Physics and Astronomy Research Council (PPARC), and Italy's National Institute for Research in Nuclear and Subnuclear Physics (INFN), along with 15 other associate organizations, including IBM. EU investment is likely to be announced at a summit of political leaders in Lisbon next month.

The call for a proposal — unusual in that it asks for a single application from the European research community — represents a green light for the Géant project. This has been proposed by Dante, the Cambridge-based coordinator of 31 national research networks in Europe, as the successor to the current Ten-155 European research network.

Géant would boost the network's capacity from an average of 155 megabits per second to 2.5 gigabits per second this year, and to 10 gigabits per second and higher within four years. Member states fund the pan-European research network jointly, so the upgrade “will happen fast and be focused,” says one commission official in Brussels.

The new commission funding stipulates that the network must operate at the same speed throughout Europe, even in poorer countries. “There is a need to avoid partitioning the European research infrastructure fabric and weakening European integration,” says one commission document.

Dai Davies, general manager of Dante, is optimistic that this requirement can be met. “While national infrastructures are not of the same quality, this can be offset somewhat by balancing costs so that a part of the total are shared irrespective of geography,” he says.

Further support for research computing infrastructure may come from the Lisbon summit, which is likely to approve a broad commission proposal to boost the knowledge-based economy known as eEurope.

One commission official says that eEurope will extend Géant by organizing gigabit-per-second connections to the network from all campuses and research institutes in Europe. “The objective is to build a completely optical network throughout all Europe, including poorer countries, before 2005.”

EU funding is likely to be of the order of 100 million euros. The overall cost could be over 1 billion euros, and would require national and regional agreement. The European Investment Bank — an autonomous body set up to further European integration by promoting EU economic policies — is said to be interested. The EU is likely to approve spending a further 50 million euros to build fast connections linking Géant to networks across the world.

Liberalization of EU telecommunications markets has prompted a 20-fold reduction in costs over the past two years, according to Davies. He adds that progress in extending into central and eastern Europe has been easier than in Portugal and Greece, where telecommunications operators are reluctant to abandon monopolistic tendencies.

The quality of infrastructure now in sight means Europe can start work on a more advanced notion of research computing, the so-called ‘grid’. The idea is that scientists would plug into the network and immediately have at their disposal the entire processing resources of the network.

The term ‘grid’ comes from the analogy with the electricity grid, where one can plug in an appliance and the power needed is always there. The complex handling of data and jobs between centres would be hidden from the end user by a new generation of software known as ‘middleware’.

CERN's $1.8 billion Large Hadron Collider (LHC) is an ideal test bed for the grid, as it will generate 7 petabytes (1015 bytes) of data every year when it comes online in 2005. This information — requiring a thousand times more computing power than CERN can currently deliver — must be delivered to thousands of users in more than 40 countries. Using the grid concept, the Internet would function like a single computer and database rolled into one (see Nature 404, 213; 2000).

Political support for the programme is already high. Britain has pledged £100 million to grid computing. John Taylor, director general of the UK research councils, told research ministers in Lisbon last month that the grid represented a fundamental shift in computing, and would allow global teams of scientists to access large-scale computing resources and data collections distributed across many institutions and countries.

A $500 million five-year grid effort is already under way in the United States, involving 50 research centres coordinated by the National Computational Science Alliance.

Preliminary discussions suggest that Europe's grid effort is likely to “match or exceed” that of the United States, says Robin Middleton, a scientist at the UK Rutherford Appleton Laboratory and the PPARC representative of the CERN proposal.

The CERN project is seen as a forerunner for similar proposals from other disciplines with similar computing needs, such as bioinformatics and astronomy.