CSIRO's new AU$4m supercomputer will be something to Bragg about

The Commonwealth Scientific and Industrial Research Organisation (CSIRO) has kicked off the hunt for a new supercomputer to replace its existing Bragg accelerator cluster, a system the organisation currently uses to solve big data challenges in fields such as bioscience, image analysis, fluid dynamics modelling, and environmental science.

According to CSIRO's acting deputy chief information officer for scientific computing Angus Macoustra, Bragg's replacement will be capable of "petaflop" speeds to support the broad range of projects undertaken by the organisation's researchers.

Bragg is accessible to any of the research projects within the organisation, with Macoustra noting the CSIRO also makes the system available to external partners on request.

"There is a pretty diverse range of projects that the computer is expected to support. Materials research do a lot of GPU-based computing, similarly climate modelling, chemical screening, pharmaceutical screening -- there's a whole variety, a multitude of applications," Macoustra told ZDNet.

"As an example, with the current Bragg, we actually provide access through to the Cancer CRC in partnership with one of our science business units for the screening of potential cancer therapeutic treatments."

It is expected the cluster will power a new generation of groundbreaking scientific research, including data analysis, modelling, and simulation in a variety of science domains, such as biophysics, material science, molecular modelling, marine science, geochemical modelling, computational fluid dynamics, artificial intelligence, and data analytics using deep learning.

The CSIRO's tender calls for a "heterogeneous" system that combines traditional central processing units (CPUs) with coprocessors to accelerate both the machine's performance and energy efficiency.

Sticking within the organisation's AU$4 million budget, the new system is expected to significantly exceed the existing computer's performance.

While the CSIRO has not specified the number of compute nodes it is after, it anticipates the cluster to be housed in five racks of 48U, with each node including at least two CPUs of either a Power or Intel x86-64 bit architecture, with a minimum of 4 cores.

Each node will include at least two Nvidia Pascal GPUs and the nodes will connect to an existing FDR Infiniband interconnect.

The CSIRO also anticipates the cluster will be able to achieve around 1 petaflop of performance.

The existing cluster was named after Australia's first Nobel prize-winners, father-and-son physicists Lawrence and Henry Bragg. It was the first of its kind in Australia combining traditional CPUs with GPUs when it launched in 2009.

The current Bragg debuted in the Top500 list of supercomputers in the world in 2012 at number 156, then in 2014 it achieved number 7 on the Green500 -- a ranking of energy efficiency of the world's supercomputers.

Macoustra said the new system will reside within CSIRO's Canberra datacentre, hosted there along with the organisation's other HPC systems.

"All of these models use a lot of data, so we have quite a significant data storage cloud also in the datacentre," he added.

The CSIRO went to tender to find a new Advanced Technology Cluster (ATC) to replace the decommissioned Fornax system at the Pawsey Supercomputing Centre in Perth, a national supercomputing joint venture between the CSIRO, Curtin University, Edith Cowan University, Murdoch University, and the University of Western Australia.

With a budget of AU$1.5 million, the CSIRO specified the new ATC was to meet the needs of the radio astronomy research community and high-end researchers in other areas of computational science, such as geosciences, nanotechnology, and biotechnology.

In March, the CSIRO welcomed the Dell-powered Pearcey supercomputer to its Canberra site to support research in areas such bioinformatics, fluid dynamics, and materials science.

A month prior, Monash University received a M3 high performance supercomputer upgrade, using Dell's super compute platform and powered by GPU giant Nvidia.

Similarly, the Faculty of Science at the University of Western Australia also welcomed its own high-performance computing cluster to its Perth campus to assist with computational chemistry, biology, and physics.

The CSIRO's latest tender will close on Monday, December 19, 2016 and the winning system is expected to be up and running during the first half of 2017.

"There are a number of providers around Australia with proven credentials around this, and we also have a quite capable team internally -- we've been doing this sort of thing since 2009," Macoustra said, adding that he believes the team are quite capable of having the new Bragg system ready in such short amount of time.

Last week, the National Computational Infrastructure (NCI) announced that Xenon Systems will be supplying a Lenovo NeXtScale system as an extension of Raijin, NCI's current peak facility.

The new Lenovo system is expected to help NCI to meet the demand of high-performance computing by providing a 40 percent increase in capacity to its facility.

The NCI expects the Lenovo system will complement the wide range of computational resources available through NCI and will provide a major boost in capability for users with high memory requirements.

Key specifications for the Lenovo NeXtScale system, which will be integrated with Raijin's fast filesystems, include: 22,792 Intel Xeon Broadwell 2690v4 processors; 144 terabytes of memory, including 10 one-terabyte nodes; and Mellanox EDR 100Gbit/s Infiniband interconnect, configured as a "fat tree".