Two years of money does not compute
Insecure funding for IT hardware and people is frustrating scientific research in Australia, writes Trevor Clarke.
A Nobel prize-winning scientist has been joined by the former chairman of the Australian eResearch Infrastructure Council and industry chief executives in an impassioned call for long-term funding assurance for Australia's high-performance computing infrastructure.
Distinguished professor and ARC laureate professor of astronomy - and 2011 winner of the Nobel prize in physics - Brian Schmidt said the supercomputers located at universities and other facilities around the country were critical to scientific research.
"The infrastructure underpins most of the work we do in science and, quite frankly, the whole research and development parts of what Australia wants to do," he said. "The difficulty we have is everything we do in this sector has a very long time horizon. So, if we blow it today, we don't pay the price as a nation until 10 or 20 years down the track."
Schmidt said that while planning for information technology infrastructure to support scientific research should be done at least five years out, the current funding approach, which offers only two years of assurance, was a "Band-Aid".
"That is a terrible place to be because we plan and we have to plan for success. And I am afraid planning for success over the past couple of years has been a little frustrating," he said.
Queensland University of Technology deputy vice-chancellor and the former chairman of the Australian eResearch Infrastructure Council, Professor Tom Cochrane, agrees with Professor Schmidt.
"The need for a more sustained long-term approach, perhaps with the same level of assurance as we have for the more general requirement to publicly fund research, is required," Professor Cochrane said.
"Making this point does not in any way detract from the significant achievements of the period 2006 to 2013, but we need to understand that reliance on this infrastructure is an integral part of Australia's future research and we must have a more sustained way of supporting it."
In November last year, Australia achieved its best global supercomputer ranking with a powerful new Fujitsu Primergy system at the National Computational Infrastructure facility at the Australian National University, debuting at No.24 in the biannual Top 500 supercomputer rankings. It is now number 27.
NCI director Professor Lindsay Botten said the challenge was the recurrent cost of people and power. In the NCI's case, these costs can be $11 million to $12 million a year.
However, the real risk of not creating certainty in funding Australia's supercomputing infrastructure over the long term was losing talent, according to Intersect CEO Dr Ian Gibson.
"The level of capability is perhaps even more sensitive to this issue of short-term funding than the infrastructure," Dr Gibson said.
"A peak facility [such as China's Tianhe-2, the fastest in the world] may last you three or four years if you squeeze the last juice out of it.
"But it takes at least five years to build a world-class capability. If you only have two years of funding going forward you are not going to be able to attract and retain world-class capability."