Australia’s CSIRO, the Pawsey Supercomputing Centre and the International Centre for Radio Astronomy Research (ICRAR) are part of an international consortium that has completed a five-year effort to design supercomputing infrastructure that will help process the masses of data generated by the Square Kilometre Array (SKA) project.
SKA will be the world’s largest radio telescope, with ground infrastructure hosted across two nations: Australia and South Africa. Australia will host ‘Ska-Low’ while South Africa will be home to ‘Ska-Mid’. The Australian portion of SKA will comprise 132,000 low-frequency antennas.
A consortium led by the UK’s University of Cambridge and with participation from almost 40 research institutions across 11 countries, including the New Zealand Alliance (NZA) and Victoria University of Wellington, late last week announced it had completed engineering design work on the Science Data Processor (SDP) component of SKA.
Two supercomputers are at the heart of the SDP, which is the second stage of data processing following correlation and beamforming, which is performed by the Central Signal Processor. CSP design work was completed in February.
SDP takes the data from the CSP and turns it into “science data products,” the SKA Organisation said.
“SDP is where data becomes information” data centre scientist for the SKA Organisation, Rosie Bolton, said in a statement. “This is where we start making sense of the data and produce detailed astronomical images of the sky.”
One SDP supercomputer will be located in Perth and the other in Cape Town, South Africa.
SKA Organisation SDP project manager, Maurizio Miccolis, said that the SDP is estimated to have overall compute power of around 250PFlops, 25 per cent faster than IBM’s Summit — currently the fastest supercomputer in the world.
“In total, up to 600PB of data will be distributed around the world every year from SDP – that’s enough to fill more than a million average laptops,” Miccolis said.
The Perth-based ICRAR developed the execution framework for the SDP.
“Our job has been to create something called the ‘execution framework’, which will run up to 100 million tasks every few hours as it ingests and processes data flowing from the telescope at enormous speeds,” said Professor Andreas Wicenec, who heads the Data Intensive Astronomy program at ICRAR.
Wicenec said that because of the quantity of data flowing into SDP — around 5 terabits per second — “it will need to make decisions on its own in almost real-time about what is noise and what is worthwhile data to keep.”
The CSIRO said it had contributed expertise in software architecture and imaging algorithm development to the SDP consortium, drawing on the organisation’s experience with the ASKAP (Australian Square Kilometre Array Pathfinder) telescope.
The Pawsey Supercomputing Centre worked to prototype and benchmark highly parallel software, as well as contributing its experience running a large-scale data centre. The centre also helped develop the local infrastructure interface specification between the SDP and the hosting data centres.
In February, the SKA Organisation revealed that two consortia, Infrastructure Australia (INAU) and Infrastructure South Africa (INSA) — had completed their design work for the telescope’s ground infrastructure. The CSIRO led INAU, working with Aurecon engineers to tackle a range of novel problems, including the need to minimise potential radio interference from computing and power systems.