‘Graham’ brings superpower to UW
WATERLOO — In the basement of the Mathematics and Computer Building on the University of Waterloo’s main campus is the biggest supercomputer at any post-secondary institution in the country.
“Today is a day of celebration,” said John Thompson, the university’s associate vicepresident of research, in a presentation Friday at the public launch.
It is called Graham, after J. Wesley Graham, a math professor who is known as the father
of computer science at the University of Waterloo.
This $17-million machine is one of four supercomputers and data centres that will be located at four universities across Canada. Together, they form a $71 million advanced research computing network that will provide unheard of amounts of storage and unprecedented computational power to 11,000 academic researchers and their industry collaborators.
The Graham Cluster, as it is called, is made up of several rows of black, metal cabinets that hold 35,000 central processing units — or CPUs — and 1,000 servers. A typical notebook computer has four CPUs. It weighs 16,000 kilograms, and uses 650 kilowatts of power. The amount of data that can be stored is mind boggling, 50 petabytes or 50 million gigabytes.
For perspective, one gigabyte is all the information on printed paper that could fill a pickup truck. So Graham can store the information on the paper that fills 50 million pickup trucks.
It can perform more simultaneous calculations than any other Canadian academic supercomputer. That’s exactly what Duane Cronin has been waiting for.
Cronin, a professor in mechanical engineering and head of the university’s centre for automotive research, looks at Graham and sees nothing but possibilities.
Cronin builds incredibly detailed models of human bodies on computers, made up of two million different pieces, or blocks. The virtual humans are placed inside virtual cars, which are put through simulated accidents.
A typical car accident happens in thousandths of a second. Every one of the two million blocks in the virtual human must be studied through each stage of the accident. That generates so much data it takes today’s computers seven days to process the information.
Graham is so powerful it will reduce the time for modelling accidents and analyzing the data to one or two days.
“This is providing new insights into crash safety that have never been available before,” said Cronin.
Cronin focuses on what happens to vertebrae and neck ligaments in car crashes. And the computational muscle of Graham will lead to safety improvements in cars, said Cronin.
A strong breeze constantly blows through the room thanks to a powerful ventilation system that cools the supercomputer. People often wear ear plugs here, and Cronin has to speak loudly to be heard.
“This is so exciting for me,” Cronin said. “To stare at these rows of cabinets and see all of these lights blinking on and off, I can visualize what is going on inside, and it is fantastic.”
Nearby is another excited researcher, Scott Hopkins, a professor of chemistry and nanotechnology. His research is focused on determining the properties of atoms and molecules inside human bodies. What happens when they are dissolved in blood or water? How quickly do they move through cell walls?
He wants to understand how the properties of these atoms and molecules change. He wants to take advantage of those properties to produce new materials and new drugs.
The human body is complex and dynamic. Tracking how a new drug behaves when surrounded by water molecules, ions and cell membranes is an incredible problem because of the huge amounts of data that must be collected and analyzed, said Hopkins.
“Graham will deliver high performance computing power, and be able to handle a lot of big data,” said Hopkins.
In the earth sciences, high performance computers are needed for modelling changes in the atmosphere, climates and oceans.
The powerful machines will also advance understanding of quantum theory — or the behaviour of subatomic particles — and cosmology, the science of the stars, planets and universe. “These are our two book ends for our current understanding of the universe,” said Hopkins.
Ottawa, Queen’s Park, Compute Canada and the Canada Foundation for Innovation are all partners in this national network of supercomputers. The machines will be located at University of Waterloo, the University of Toronto, Simon Fraser and the University of Victoria.
“Today is being celebrated for one simple reason: Canadian researchers are getting access to much needed technology that will change our world,” said Niza Ladak, the president and chief executive officer of Compute Ontario.
Graham was built with funding from Canada Foundation for Innovation and the Ontario government.
Researchers today generate massive amounts of data that require high-performance computers for processing, said Manon Harvey, a vice-president of the Canadian Foundation for Innovation.
“Graham is the third of four national systems to be fully deployed on universities across the country,” said Harvey.