If space is the last frontier of science, a group of local astronomers is hoping its exploration will be advanced with computing technology developed firmly on Earth as the electronic Very Long Baseline Interferometry (eVLBI) project gets under way this month.
Led out of Swinburne University of Technology’s astronomy department, the eVLBI project deployed a third cluster at Narrabri, NSW, last month which will work with existing supercomputers at Swinburne in Victoria and Parkes in NSW. VLBI involves combining signals from two separate locations to synthesise a telescope that is the size of the baseline.
The project’s goal is to demonstrate that Western Australia should be chosen as the site for the world-wide Square Kilometre Array (SKA) telescope with an area 30 times greater than the largest so far built.
Swinburne astronomer Steve Ord said the SKA telescope is a huge multinational project that will revolutionise astronomy.
“A site hasn’t been decided but it’s down to America and Australia and we have the best chance of getting the site,” Ord said. “If we win it will involve building hundreds of sites scattered around WA.”
To estimate how much computer power will be required to process vast amount of data acquired by such a telescope, Ord put it into perspective. “To be able process the data from the six radio telescopes in Australia, we require around 100 gigaflops, provided that the systems are optimised,” he said. “For a telescope such as the SKA we would be looking at computing power in excess of 2 petaflops!”
So far clusters have been installed in three sites, but Ord said the plan is for six locations with a separate cluster in each, with one likely to be in WA.
Craig West, eVLBI group research assistant, said in order to process the vast amounts of data that are generated at multiple locations around the country many computers could be placed at each site, or one larger computer placed at a "central" location with all data being sent there.
“To be able to send the data we require very fast, and dedicated, links with speeds of 1Gbps at a minimum. However, this still leaves the problem of recording the data, and sometimes we simply can't transmit all the data,” West said.
“We can use storage at each location and send the data home in the back of a station wagon. Or we can put limited processing systems at each site for help in processing the data prior to sending it, or they can used as a larger part of big cluster of computers – grid processing.”
In order to win the SKA telescope’s location, the researchers believe the two most important aspects are to have the best site and engineering leadership.
“The proposed site in WA is ideal as there is little radio interference,” West said. “Displaying the engineering know-how [is important]. This will be a massive international collaboration - but many of the technologies simply don't exist yet – we are displaying an eagerness to solve a lot of the technological problems, and provide solutions that can make this project happen.”
Ord said the software that will run over the grid is being developed now and will essentially process data from the radio telescopes.
“The main aims are VLBI processing in near-real-time, as opposed to the months that it can take presently,” Ord said. “Grid computing and fast networking links allow us to distribute the data between different sites for processing.”
The radio telescopes owned by the CSIRO ATNF have 512kbps second connections but the astronomers hope the AARNET3 regional network will be able to supply 1Gbps by the end of the year with possible expansion to 10Gbps.