Grand challenges of high performance computing
[img_inline align=”right” src=”http://padnws01.mcmaster.ca/images/Hahn_Mark_Requin.jpg” caption=”Mark Hahn, SHARCNET system administrator, stands next to one cabinet of processors. Photo credit: Rob Tatlock”]Simply put, David Earn wants to help eradicate infectious diseases. Stephen Tullis wants to improve the stability of flames in gas turbines to help decrease pollution. James Wadsley wants to understand the nature and origin of galaxies and Nikolas Provatas wants to produce stronger, yet lighter materials that might one day lead to safer cars.
These diverse research projects share the same vital component: research results achieved through the use of high performance computers (HPC).
The Shared Hierarchical Academic Research Computing Network (SHARCNET) — Canada's largest HPC consortium — formally launched its second phase of development, celebrating funding totaling more than $100-million, new equipment and new partners.
McMaster's newest high performance computer cluster 'Requin' — French for shark — will be up and running next week, a parallel cluster comparable to 1536 high-powered desktops, if each possessed more than four times the normal amount of memory. Requin is capable of consuming and producing massive amounts of data in a very short period of time and will likely rank in the top 100 of the Top 500 List of Supercomputer Sites in the world.
McMaster University is one of the original four universities and two colleges that formed the SHARCNET consortium in 2001, now expanded to 16 universities and colleges forming a 'cluster of clusters' of high performance computers in south-central Ontario.
“Our first phase brought people and technology together and we're now beginning to see the development of some tremendously exciting projects,' says Hugh Couchman, professor of physics and astronomy and SHARCNET's Scientific Director. “HPC of this scope provides a fundamentally different level of capacity by enabling a wide range of very complex computations across many locations tightly coupled by a dedicated, high performance network.”
James Wadsley, a theoretical astrophysicist and assistant professor of physics and astronomy, notes that, “with Requin, researchers will be able to conduct Grand Challenge Projects, solving problems that can't be solved in a reasonable amount of time with today's computers, because the scope of the project is just too large.”
Wadsley has used SHARCNET to create simulations of stars crashing together, allowing him to further explore the complicated system of galaxies that astronomers can't see, but can infer from the models they have created using computational research.
The added bonus of SHARCNET and having facilities such as Requin on campus is that scientists the world over — those who are “dreaming big” as Wadsley puts it, those who want to be “real players on a world class stage,” will be attracted to the very real benefits that our high performance computers can offer.
As an example of this, Wadsley cites the master's student joining his research group this summer. A recipient of the top Natural Sciences and Engineering Research Council (NSERC) scholarship, the student chose this University because of the opportunity to access SHARCNET for his research.
Similarly, Stephen Tullis, an assistant professor of mechanical engineering chose to come to McMaster from his native England and a position at Cambridge University to access the high performance computing time and capabilities he needs to pursue his research in turbulence and combustion. Using SHARCNET, he is able to model “flames in a box” — duplicate the sorts of flames you'd find in gas turbines that power generators — and design out the instabilities that ultimately produce pollutants such as nitrous oxides.
Materials scientist Nikolas Provatas is able to research and pattern at small scales to represent the large scale properties of aluminum. High performance computers provide him with the opportunity to perform simulations that are relatively cheap and quick and can ultimately save industry millions of dollars in testing costs.
The infectious disease research of mathematician David Earn is an example of a Grand Challenge Project. Earn, whose specialty is mathematical biology, is trying to understand the path of epidemics and by doing so, will provide the data needed to improve vaccination strategies to combat infectious diseases such as influenza.
“The flu kills more than 30,000 people a year in Canada. If we can understand the ecology and evolution of this disease, we can start thinking about response strategies — design, develop and target vaccines more effectively,” says Earn.
Funding totalling more than $100 million from the federal government, through the Canada Foundation for Innovation (CFI), and the province, through the Ontario Innovation Trust (OIT) and Ontario Research and Development Challenge Fund (ORDCF), as well as private sector and industry contributions, has enabled the construction of SHARCNET's leading computing facilities and the recruitment of top scientists to the province. McMaster's Requin cluster is now the most expensive in SHARCNET, representing an almost $16 million investment.
