Pawsey brings new $2m GPU cluster online – Hardware
MWA antennas. Supply: Pawsey Supercomputing Centre
The Pawsey Supercomputing Centre has set stay a new $2 million cluster of graphics processing models (GPUs) to electric power investigate making use of the Murchison Widefield Array (MWA) telescope.
The cluster was procured from HPE before this yr as aspect of Pawsey’s $70 million cash refresh job funded by the federal government.
Dubbed Garrawarla, which means ‘spider’ in the language of the Wajarri people whose land the telescope sits on, the dedicated seventy eight-node cluster processes the petabytes of details streamed from the MWA.
GPUs ended up chosen to provide extra electric power to the astronomy investigate that makes use of compute-intensive artificial intelligence and equipment understanding workflows for details analytics.
Researchers who experienced early access to the cluster even though it was in pre-output noted application performing two to 8 instances a lot quicker than on its predecessor.
MWA director, Professor Melanie Johnston-Hollitt, stated researchers from twenty organisations based mostly in five unique countries are benefitting from the Garrawarla installation.
“The MWA collects a petabyte of details a thirty day period, and we’ve by now archived about 30 petabytes at Pawsey,” Johnston-Hollit stated.
“Having a dedicated program signifies that we can perform with Pawsey to great-tune the program and maximise the performance of our workflows.
“Garrawarla and Pawsey’s know-how will accelerate our investigation and processing of this details, assisting us to expose additional about the origins and construction of the universe.”
Pawsey employees are now setting up the application natural environment to enable additional MWA investigate to migrate workflows onto Garrawarla.
Pawsey has also signed a separate $1 million deal with HPE to procure ingest nodes to support researchers making use of CSIRO’s Australian Sq. Kilometre Array Pathfinder (ASKAP) telescope.
The supercomputing centre described the ingest nodes as “one of the most essential components” of the pipeline in between the telescope and the Pawsey details keep, which houses ASKAP’s final details merchandise.
The nodes acquire the details in genuine time from the correlators at the Murchison Radio-astronomy Observatory (MRO) and writing the details to disk to be processed on Pawsey’s Galaxy supercomputer.
The new ingest nodes will characteristic the most current AMD processors, giving researchers two times as substantially bandwidth to and from the CPU as the recent era of nodes, as effectively as additional memory channels.
Pawsey’s executive director Mark Stickells stated the new infrastructure will assist the centre’s techniques maintain up with the deluge of details that are expected to be created when ASKAP starts total sky surveys in the coming months.
“The ASKAP ingest nodes will boost the capacity to transfer details from the MRO into Pawsey’s storage techniques, and will assist to unlock the price of the astronomical details produced by the telescope,” he stated.