An "infinite" demand for Australia’s supercomputing resources could threaten the quality of climate change science if more high performance computing isn’t found, a leading academic in the field has warned.
According to professor Dave Griggs, director of the Monash Sustainability Institute (MSI) and CEO of ClimateWorks Australia, the demand for supercomputing resources to run complex climate change modelling poses a challenge for scientists.
“We have an infinite demand for supercomputing and the quality of the predictions you can make are a combination of how good you are at the science you put into the models, and then how powerful the supercomputing you have got to run it on,” he said.
“You need to competitive in both of those things. There is no point in putting bad science into the best computer in the world. Equally, there is no point in putting good science into a computer that is not competitive.
“If Australia were to fall behind, significantly, its competitors, then that would threaten the quality of the science. That hasn’t happened. You don’t have to have the biggest computer in the world to be the best — because the quality of science is important. But, if you are so far behind everybody else [in supercomputing] … then it would be impossible maintain your competitiveness with colleagues in other countries.”
Offering a sense of climate change science’s reliance on supercomputing to run modelling, Griggs said one model currently in use “is probably one of the most complex programs ever written… and it takes about six months to run on some of the most powerful supercomputers in the world.” “These models require a huge amount of supercomputing power; they are not trivial exercises.”
Demand for supercomputer access is not only coming from the climate science community. Australia’s bid for the $2.1 billion Square Kilometre Array (SKA) radio telescope is a major driver behind the creation of the Pawsey high performance computing centre.
In fact, demand for processing power has driven key players behind the SKA to explore the use of grid-based Cloud computing, with the aim of potentially harvesting the computing and storage power of desktops worldwide.
In July Swinburne University of Technology moved ahead with its plans for the rollout of a hybrid CPU- and GPU-based supercomputer, signing Silicon Graphics to provide the first phase of the project.