The University of Waterloo’s mathematics faculty is hoping to shed new insights on everything from climate change to the effect of pollution, thanks to an upgraded computing architecture.
The faculty has significantly upgraded two multi-processor Altix systems that it purchased from SGI, and is using the systems to more accurately model fluid dynamics in lakes and oceans.
The faculty has been using Altix 350 and Altix 3700 systems since 2004 and has upgraded them multiple times, most recently in September. The Altix 350 machine, which originally shipped with two Itanium processors and 12GB RAM, has been upgraded to 16 Itaniums and 96GB RAM. The extra processing power will let researchers explore the mixing of water within lakes and oceans in more depth, explains software specialist Robyn Landers, who works in the faculty administering Unix systems.
“This has to do with how waves form and travel, and go across the ocean,” he said. “How those topographical features interfere and change the motion of the wave, and how it bounces back and mixes the water around.”
This area of study has multi-disciplinary uses. In Lake Erie, for example, oxygen levels have been dropping dramatically at the end of the summer, which can have an effect on the wildlife there. Understanding how water of different temperatures mixes together can shed new light on the problem. “The kind of work our researchers do can help biologists understand what’s happening with algae and fish,” Landers said.
Similarly, fluids research can provide insights into the wider relationship between the movement of ocean water and climate change. Even the navy is interested in this area of research to help understand the propagation of sonar sound through water, which critics have said can have an effect on sea life.
The SGI machine uses a shared memory architecture, which presents a system of memory distributed around different processors as though it is a single piece of memory. This can make it easier to run older program code because it doesn’t have to be rewritten for a clustered architecture, in which memory is presented to programmers as distributed across many different machines.
The ocean wave modeling code used by Kevin Lamb, the chair of the University’s applied mathematics department, is 15 years old and was written to work within a single block of memory. Using a shared memory architecture means that the code doesn’t have to be rewritten to work in a clustered environment. “What we typically have is code that’s been around for 15-20 years and it’s a lot of work to get them over to that cluster,” explained Michael Brown, sciences market segment manager at SGI.
The SGI machines used by the faculty also have enhanced numerical algorithms which have made processing faster in applications such as simulation, explained Landers. The result will be a new understanding of the way oceans and lakes work. “This reveals more subtlety and more detail in the interaction of the waves and mixing of the water, and provides more information.”
Such enhancements enable researchers to run processing jobs more quickly and cover broader geographical areas. They can also begin acknowledging the effect of seabed topology as it creates more complex, higher activity areas of fluid movement.
The Altix 3700 machine now has 64 processors and 192GB RAM, which Landers says enables other groups in the faculty to run simulations that were not previously possible. These groups are conducting research into areas including number theory in pure mathematics, and optimizing algorithms in applications such as computer science and manufacturing.