December 2011 - Issue #14
Swinburne’s gSTAR heralds ‘mega science’
Story by Clarisa Collis
- A new supercomputer is helping to rapidly process huge amounts of astronomical data.
- It draws on the computer gaming industry’s development of GPUs for rapid graphics display.
- gSTAR is the first large-scale GPU supercomputer available to Australia’s astronomy community, and is set to broaden the scope of scientific exploration.
The search that culminated in the discovery of a diamond planet in our Milky Way galaxy earlier this year came after researchers sifted through more than a thousand billion bytes of coded data. It took three years of data processing using Swinburne University of Technology’s 12-teraflop supercomputer, the Green Machine.
Now, a new 120-teraflop supercomputer driven by graphic processing units (GPUs) is set to accelerate the rate of science and knowledge discovery at Swinburne, crunching data 10 times faster than the Green Machine. It means the equivalent of the diamond planet search could potentially be done in a week.
The massive jump in capability has involved the investment of more than $1.5 million by Swinburne and $1 million from the Australian Government’s Education Investment Fund.
Swinburne Pro-Vice-Chancellor for Research Matthew Bailes says the new GPU Supercomputer for Theoretical Astrophysics Research (gSTAR) is among the top six supercomputers in Australia, and will probably rank in the world’s top 200 machines.
Professor Bailes says the surge in processing power since Swinburne invested $1 million in the Green Machine in 2007 shows how quickly the power balance shifts between supercomputers.“gSTAR can perform more operations in a few seconds than the earliest supercomputers could conduct over their lifetime,” he says.
Professor Bailes anticipates gSTAR’s computing power and its data-throughput capacity will continue to increase through further technical developments in the next two years, with upgrades to the system potentially doubling the current one petabyte or 1.8 million gigabytes of disk space.
The system is supported by a quad data rate (QDR) InfiniBand network that transfers data to GPUs and large memory nodes, each with 512 gigabytes of memory, at 1000 times the speed of wireless internet connection.
Game power builds cosmic picture
Contained in 11 cabinets on the sixth floor of the Advanced Technologies Centre, the machine will meet the supercomputing needs of three Swinburne faculties: Information and Communication Technologies, Engineering and Industrial Sciences, and Life and Social Sciences.
Swinburne Supercomputer Manager Associate Professor Jarrod Hurley says that GPUs – developed to capture the enormous visual and simulation processes in computer gaming – add another dimension to astronomy.
In contrast to the traditional architecture of supercomputers, which rely on central processing units (CPUs) to process information in sequence with an emphasis on control logic and fast memory, GPUs perform numerous high-speed tasks simultaneously. Aside from immense processing power, they also add exciting new visual or graphical details to endless strings of numeric figures.
Professor Warrick Couch, Director of the Swinburne Centre for Astrophysics and Supercomputing, says gSTAR allows researchers to create better maps of the universe.
“gSTAR will show all the galaxies are clumped, and in doing so will reveal a new microscopic context that provides a detailed perspective on objects within it, such as individual stars,” Professor Couch says.
As astronomers double the amount of information needing to be processed every few years, gSTAR is geared to keep pace with this increasing demand by facilitating more than 100 trillion computations per second, while using less hardware or nodes.
On the cusp of a new generation
Suited to “mega-science projects”, gSTAR is poised to receive a “deluge of data” from an emerging generation of telescopes equipped to generate a million billion bytes of information in a single night, Professor Couch says.
“GPUs will make a major contribution to processing data from new optical and radio telescopes, like SkyMapper in NSW and the Australian Square Kilometre Array Pathfinder in Western Australia, that are designed to image huge amounts of the sky,” he says.
Professor Couch speculates the step up in supercomputing could also help resolve a mystery that defies the laws and theories of physics: the nature of dark energy – a force that is inexplicably causing the universe to expand at an accelerating rate.
A second phase of the new supercomputer that centres on CPU-based hardware will free up Swinburne’s superseded Green Machine for other tasks, such as three-dimensional animations, including plans for a high-resolution astronomy movie for IMAX. This second phase is due for completion in mid 2012.
Meanwhile, Professor Bailes says although Swinburne tends to renew its supercomputer every two to four years, the onsite facility may one day be replaced with a ‘supercomputing cloud’ to provide a massive globally shared resource.