Australia's supercomputers need 'urgent' refresh, or life-saving research at risk

The whitefly is having a devastating effect across East Africa by spreading viruses to the cassava plant, a staple food in the region. An infestation can decimate an entire harvest, leading to economic hardship and famine.

While they look identical, there are actually many species of whitefly. In order to develop resistant crops, they must be distinguished and the best way to do this is to examine their genetic differences.

Professor Laura Boykin and her colleagues at the University of Western Australia have been collecting whiteflies, extracting the insect DNA, and using sequencing machines to generate their genome.

To make sense of the billions of genetic base pairs, researchers have to apply sophisticated algorithms, to genetically distinguish the flies. That process takes significant computational resource. It takes a supercomputer like Magnus.

Magnus is Pawsey Supercomputing Centre in Perth’s flagship machine, and one of the most powerful supercomputers in Australia. The most powerful is Raijin, operated by National Computational Infrastructure (NCI) at the Australian National University.

But Australia’s two Tier 1 high performance computing (HPC) research facilities are in “urgent” need of a refresh. Their supercomputers are fast approaching end-of-life, just as demand from academics needing the compute power to crunch massive datasets is rising.

It’s not hyberbolic – given research such as Boykins’ – to say there are lives at stake.

"HPC and genomics is the future," Boykin told Computerworld. "Building capacity at the integration of these two disciples is key to solving huge problems – like famine in East Africa."

Rank slip

There comes a point in every supercomputer’s lifecycle when they are not so super anymore.

When it became fully operational in 2014, Pawsey’s petascale system Magnus was the 41st most powerful computer in the world.

By 2015 it had slipped to 58 on the LINPACK Benchmark Top500 ranking. The latest list, released last month, placed Magnus at 111.

“The problem is compute technology – in many ways it accelerates faster than you'd like it to!” says David Schibeci, head of supercomputing at Pawsey. “At a certain point, we understand that all supercomputers have an end of life.”

The support contract for Magnus – which is currently running at near capacity – lapses in September. Pawsey has received a government funding boost to cover its next two years of operations.

“Part of that money will be to extend that support contract so we can still look after Magnus,” Schibeci explains.

The NCI’s Raijin supercomputer currently comes in at number 70 in the world ranking, with a performance of 1.67 Petaflops – comparable to about 40,000 desktop computers working simultaneously. When it debuted in 2012, Raijin was placed at 24.

The current ranking is only thanks to an upgrade last year – paid for by an emergency government ‘Agility Fund’ – which helped Raijin rise from 121st place in 2016.

“However, despite this upgrade, NCI’s core computer is fast approaching the end of its serviceable life and requires urgent replacement,” the NCI said in a statement last month.

Without upgrades, vital research cannot happen, says the NCI.

“Without a replacement for NCI’s current computer, the competitiveness of Australia’s research will suffer and much important research will simply not be possible,” the organisation said.

It is not so much the hardware that is important, but the research it enables, says Pawsey's Schibeci.

“What is a supercomputer?" he says. "It is a very power hungry, very noisy piece of scientific equipment that allows researchers to answer their questions quicker, so they can ask the next question and the next question after that."

Page Break

World saving work

Between them, Pawsey and the NCI have assisted in research to detect gravitational waves, develop disease resistant wheat, retrieve newly fallen meteorites, map the universe in three dimensions and model climate change across millennia. The facilities will also play a role in the Square Kilometre Array global project to investigate five fundamental questions about the universe.

“It’s very exciting. I'm a computer nerd from way back. I've been allowed to look after larger and larger supercomputers, the largest in history,” says Schibeci.

“I wake up in the morning knowing that I'm looking after a service that's allowing researchers to solve some really important scientific problems. [Boykin, for example, is] trying to develop ways to try to at least control if not eradicate that transfer of the virus. At a base level I'm helping someone to try and fight world hunger.”

According to Australia’s chief scientist Dr Alan Finkel the continuation of such research is at serious risk.

In May, Finkel released the National Research Infrastructure Roadmap. In it, he said the government needed to “urgently address” the HPC situation.

“An immediate priority is the need to refresh Australia’s national HPC,” the report states. “[It] underpins the most advanced analysis and simulations in research fields, such as medical science, environmental modelling, physics and astronomy and is vital to maintaining a globally competitive research system.”

Pawsey “is eager to see this implemented” the group said. The NCI said if it’s supercomputer “is not upgraded soon Australia’s research capability will be compromised”.

Pawsey is nevertheless already planning for Magnus’ replacement. It must.

“Because it moves so fast we just need to be nimble in terms of looking at the services that we provide to Australian researchers,” Schibeci says. “And that's why we try to always look for where those opportunities lie.”

Its new Advanced Technology Cluster, Athena will inform the requirements of its next supercomputer.

Roadmap

The future of HPC in Australia is now in the hands of the government.

“The roadmap made a number of recommendations that the government will consider over the coming months,” education minister Simon Birmingham said at the time.

In the 2017-18 Budget, the government announced the development of a Research Infrastructure Investment Plan which will “inform its consideration of the 2016 Roadmap”.

The plan is being developed by the Department of Education and Training and the Department of Industry, Innovation and Science in consultation with the Commonwealth Science Council, Innovation and Science Australia and key stakeholders.

“The Plan will systematically assess the need for future research infrastructure investment, including HPC, against Government priorities and will be informed by the 2016 Roadmap and an inventory of existing national research infrastructure,” a Department of Education and Training spokesperson told Computerworld.

The importance of continued long-term investment cannot be overstated.

"We could not have done our work without access to Magnus," Boykins says. "The computations would have taken years on a normal computer. Without HPC, our work would stop."