One million hours on a super-fast computer

One million hours to tackle very big problems on a super fast computer

The University of Canterbury is helping New Zealand researchers take on the greatest challenges of their careers to date with a million free hours of use of its BlueFern supercomputer facility.

The BlueFern High Performance Computing (HPC) Grand Challenge was established to help researchers tackle some of the largest scientific problems.

BlueFern Director, Professor Tim David, says the BlueFern Grand Challenge aims to turbocharge New Zealand science.

"Not only will BlueFern be helping New Zealand researchers solve the most challenging problems, but we will also act as a pathway to overseas programs such as the American INCITE program that awards tens of millions of hours on supercomputers worldwide."

The challenge's major award has gone to a project led by Nick Golledge from the Antarctic Research Centre at Victoria University of Wellington. This project will utilise UC's supercomputer to run a 3D ice sheet model originally developed at the University of Fairbanks, Alaska, to analyse changes in New Zealand glaciers, such as the Tasman and Franz Josef. The simulations will use present day topography and climate as boundary conditions to analyse glacier growth and decay over the last 100,000 years.

Nick Golledge says: "Our project has reached a mature stage from which it could benefit greatly from higher-resolution simulations that will enable our research to produce higher-impact science."

"This has direct consequences for water resource management as well as for the tourism industry. Can we, for example, predict how much these glaciers will retreat over the next century? What might be the socio-economic impacts of these changes?"

Researchers from the University of Auckland's Computational Astronomy Group led by Philip Sharp, which has received a development award, will use simulations to study the evolution of the Solar System. The group's models consist of the Sun, planets and a large number of asteroids. It will use UC's supercomputer to model asteroid impacts on Earth.

Simon Greenhill from the Computational Evolution Group at the University of Auckland will use BlueFern to analyse a database that currently contains 680 languages to advance the group's study of linguistic and cultural change in the Pacific.

"We will use Bluefern to analyse data from the largest language family in the world using the most sophisticated models of lexical change and phylogenetic dating available. The results will generate new insights into the peopling of the Pacific and resolve longstanding debates about how and when humans entered Near Oceania and Micronesia, and help find the ultimate origins of the Austronesian peoples in South-East Asia. The resulting phylogenies will then serve as a backbone for future studies of cultural evolution."

A third development award has gone to Dr Mik Black and Dr Tony Merriman at the University of Otago who intend to use BlueFern to conduct an exhaustive search of 500,000 genetic variants to identify gene-gene interactions. This work will generate important biomedical insights into the pathways underlying human disease, providing the potential for modification by environmental changes or drug therapies.

BlueFern was established in 2006 by the University of Canterbury in collaboration with IBM to make it easy and convenient for researchers to use world class supercomputers. BlueFern features the first IBM Blue Gene to be installed in the Southern Hemisphere and an IBM p575 super cluster. The systems are made available via the KAREN advanced research network. It supports both national and international collaborators and has over 200 users.

As one of New Zealand's supercomputing and services facility for research and development BlueFern aims to help improve New Zealand's IT skills and research capability and provide infrastructure that contributes to scientific and industrial development. It continues to grow and provide a world-class environment for those wishing to get to the problems that have so far eluded scientists and engineers simply because "it was too computationally intensive".

ENDS