Bryce eager to tame ‘the beast’

WESTERN Australia is poised to become a global competitor in the world of supercomputing, which could assist the state in securing the Square Kilometre Array (SKA) project and benefit various other astronomy industries.

iVEC, a joint venture between CSIRO and WA's four public universities - Curtin University, Edith Cowan University, Murdoch University and the University of Western Australia - will build and operate the large-scale supercomputing system at Technology Park in Bentley.

Known as a petascale supercomputer, it will serve the needs of leading Australian computational researchers with a special focus on radio astronomy and have a greater capacity than that of megascale, gigascale and terrascale systems, but will not eclipse hexascale systems.

Beyond the technical jargon, the potential benefits from such a powerful supercomputer linked to Australia's eResearch Grid (incorporating two other supercomputers in Canberra and Melbourne) are incredible, particularly in supporting the state's bid to host the SKA project in the Mid West.

Substantial power is required to drive and cool the supercomputer with annual costs estimated at between $3 million and $4 million, equivalent to the power consumption of 3,000 to 4,000 homes.

iVec chairman Mal Bryce is excited about building the specifically designed structure to house the supercomputer in the next two years and then hopefully switching the supercomputer on in 2012.

"We're being funded by the federal government, so that over a four-year period there's drawdown of $20 million a year," Mr Bryce said.

"That's $80 million out of that particular super sciences program grant, so that in year four of the project, in 2012-13, what we affectionately call 'the beast' will go live.

"And we anticipate another $20 million in funding for future networking capabilities.

"It's a very exciting moment for those of us interested in scientific infrastructure, it's the biggest [science] infrastructure investment in the state's history."

Mr Bryce said industries requiring excessive computer modelling, such as nanotechnology, biotechnology, geoinformatics, climate science and atomic science, would be able to reduce modelling time from many months down to just days.

"And it has direct relevance for use in the oil and gas and other branches of the resources industry for geoinformatics and issues relating to exploration in particular," he said.