World’s biggest ‘virtual supercomputer’ given the go-ahead
The Particle Physics and Astronomy Research Council has today announced
GBP 16 million to create a massive computing Grid, equivalent to the
world’s second largest supercomputer after Japan’s Earth Simulator
computer. This Grid, known as GridPP2 will eventually form part of a
larger European Grid, to be used to process the data deluge from CERN,
the European Particle Physics Laboratory, when its new facility, the
Large Hadron Collider (LHC), comes online in 2007.
GridPP is a collaboration of UK Particle Physicists and Computer
Scientists working together to create a Grid for particle physics,
enabling them to process the vast volumes of data generated in
experiments. The LHC, a particle accelerator which will probe the
nature of matter, is expected to generate data at a rate equivalent to
20 million CDs a year.
Professor Ian Halliday, Chief Executive of PPARC, said “GridPP2 will
place UK particle physicists in a prime position to exploit physics
from the Large Hadron Collider.”
“The GridPP2 Grid will address the future computing requirements of all
the UK Particle Physics Experiments and should provide efficient
sharing of resources between Particle Physics and other disciplines at
the institutes.” added Steve Lloyd, GridPP Collaboration Board Chair.
Grid computing shares the resources of connected computers, much as the
World Wide Web (also created at CERN) enables the sharing of
information between computers. By connecting large numbers of computers
together, particle physicists are able to run simulations and analysis
in a fraction of the time it would take to run on a single machine.
Such work can also be done on supercomputers, but as these are custom
built they are expensive and in high demand. The benefit of Grid
computing is that it is constructed from cheap units and can be
expanded or reduced to fit the users’ needs.
Dr Neil Geddes, PPARC’s Director of E-Science said “Today’s money will
be used to create a grid equivalent to 20,000 1GHz personal computers.
This is the largest in the world to be funded so far.” For the past
year, GridPP have been running a prototype grid or ‘testbed’ across ten
UK sites. From this they have developed the middleware needed for a
larger Grid.
“GridPP2 will test new Grid computing technologies on a scale that we
could have barely considered two years ago.” said Tony Doyle, the
GridPP Project Leader “The Grid deployed in phase 1 taught us about the
importance of a series of testbeds where the software is incrementally
integrated and tested within an annual deployment lifecycle. Running a
stable large-scale grid service will be a major challenge and for this
reason a key component of GridPP2 will be the establishment a core
production team at the heart of deployment.”
Middleware is the programming that allows the software (the programmes
the scientists are using) to take advantage of the hardware (the
computing resources they need to access). Middleware tackles issues
such as security (e.g. allowing outside users access to a site’s
computers) and ‘brokering’ (breaking data up into packages to be sent
around the country or even world for rapid processing).
GridPP’s testbed was incorporated into the LHC Computing Grid in
September 2003, which was the first time a production grid was deployed
world-wide. GridPP is also working with projects such as the EU-funded
Enabling Grids for E-Science in Europe (EGEE) which will integrate
current national, regional and thematic Grid efforts to create a
seamless European Grid infrastructure for the support of the European
Research Area.
The experience gained in the GridPP project forms the basis of the much
wider deployment of scientific computing grids which we are seeing
across UK Universities through the UK’s e-Science programme. Industry
has also been quick to appreciate the benefits of these technologies.
