The first computer program Ryan Grant ever wrote let him move a crude rendering of a hot air balloon around the glowing blue screen of his Commodore 64.
He was five years old.
Today, nearly four decades later, the software he designs is essential for the supercomputers that are the backbone of artificial intelligence and that are used for everything from designing new drugs to forecasting weather.
The skills Dr. Grant, Sc鈥04, MSc鈥05, PhD鈥12, developed bouncing a hot air balloon around on a primitive home computer now support him in designing the software that interprets the nearly limitless complexity of our planet鈥檚 weather systems.
鈥淚 took a very keen interest in computers at a very, very young age,鈥 says Dr. Grant, an assistant professor in computer engineering with Smith Engineering at Queen鈥檚.
鈥淚 can distinctly remember going with my mother to Zellers before Christmas when I was four to buy a Commodore 64,鈥 he says. 鈥淚 loved that thing from the moment we got it.
鈥淚 think computers were a draw because the idea of controlling what was on your TV screen was just so new at that time,鈥 he says. 鈥淯p until then, electronics were really a spectator activity. Computers brought that out into something that let you be creative, play a game with friends, just do whatever you could think of. It was like Lego blocks 鈥 it鈥檚 a creative toy and a fun one at that.鈥
That love of computers still burns today, with Dr. Grant and his colleagues at Smith Engineering putting Queen鈥檚 on the map in Canada鈥檚 supercomputing industry.
Coming to Queen鈥檚 in 2021 after spending nine years working with the world鈥檚 most advanced supercomputers at Sandia National Laboratories in Albuquerque, N.M., was more than just a career change. For Dr. Grant, it was a homecoming.
Not only is he an alumnus, but he and his family have deep roots in Kingston 鈥 and at Queen鈥檚.
The connection began with Dr. Grant鈥檚 great-grandfather on his mother鈥檚 side. Alfred Plumb was a janitor at the school鈥檚 former Students鈥 Memorial Union in the 1920s.
鈥淗e worked there for 40 years. I know that because I have a silver platter that says, 鈥淎.H. Plumb, thank you for your 40 years of service,鈥 Dr. Grant says.
鈥淗e was nicknamed Sarge. He taught the students Morse code,鈥 Dr. Grant says. 鈥淗e was a First World War veteran, but sergeant wasn鈥檛 his rank. They called him that because he was a little bit ornery. He didn鈥檛 have time for nonsense.鈥
That included chastising students if he found them with their feet up on the tables, according to family lore.
Alf and his family, including daughter Gwendolyn Plumb, Dr. Grant鈥檚 grandmother, boarded Queen鈥檚 students at their house on Earl Street. Eventually, Gwen Plumb found a job at Queen鈥檚 as well, working as a maid for students in Leonard Hall.
鈥淏ack in the day, student residence rooms actually had janitorial service. They would come into your room, clean it up, and make your bed for you,鈥 Dr. Grant says.
鈥淭he students used to dissect frogs in their rooms, even though they weren鈥檛 supposed to. That would freak her out. She was not one for creepy crawlies.鈥
Dr. Grant鈥檚 mother, Judy Grant (nee Bedford) attended Queen鈥檚 for her BSc (Geography, 鈥73). Judy and Dr. Grant鈥檚 father, William Grant, settled in Ottawa鈥檚 west end, nurturing young Ryan鈥檚 interest in the new field of home computing and in his Commodore 64.
鈥淚 was actually working on it before I went to kindergarten. I鈥檇 start tapping away and put together simple programs, working from this big, gigantic instruction book. I read at an early age and it seemed simple for me. I was always tinkering with computers,鈥 he says.
鈥淭here was a really nice scene in that area of Ottawa where I grew up. All the kids in the neighbourhood had home computers and were interested in playing games on them and doing interesting things. The parents were around to help a little bit but, for the most part, all the kids got better than the parents pretty quickly.鈥
There was never any doubt that Dr. Grant would study computer engineering. Though he was offered a full scholarship to attend Carleton University, he chose Queen鈥檚.
鈥淚 wouldn鈥檛 say I chose Queen鈥檚 just because the family went here, but there was a strong pull,鈥 he says. 鈥淢ostly I wanted to try something new 鈥 outside of Ottawa.鈥
And there was another draw: he could live cheaply, staying with his grandparents, Gwen and Donald, at their home on Windsor Street, in Kingston鈥檚 Reddendale neighbourhood.
After graduating, he worked for a while with the .
鈥淚 could have gone anywhere, but Queen鈥檚 has some advantages in this space. Queen鈥檚 has some of the fastest-growing AI research programs in Canada.鈥
鈥淚 worked in an area where people had advanced degrees and all the interesting stuff seemed to be happening for people who had advanced degrees,鈥 he says.
He was soon back at Queen鈥檚 to do his master鈥檚 in computer engineering. He submitted his first paper the morning of Sept. 9, 2005. He remembers the date easily: later that afternoon he married Meagan Grant (nee Garrett, Sc鈥04 civil engineering). The couple met in a mechanical drafting lab as undergraduates.
After completing his PhD at Queen鈥檚 in 2012, Dr. Grant took jobs in the United States. He was recruited to join , where he worked with some of the largest, most powerful supercomputers in the world.
鈥淚n the States I was building this world-class stuff. I had awesome toys to play with,鈥 he says.
Compared to the United States 鈥 or indeed most countries in the world 鈥 Canada is in the minor leagues of supercomputing. Supercomputing power is measured in 鈥渇lops鈥 鈥 floating operations per second (see accompanying story) 鈥 and a rough way to rank a country鈥檚 supercomputing resources is to add up the number of petaflops in its national inventory.
Canada鈥檚 total is 41.5 petaflops (a petaflop is a million billion floating operations per second).
鈥淐anada is one-tenth the size of the United States so you鈥檇 hope that we would have one-tenth of its supercomputing power,鈥 鈥 which would put the U.S. at about 410 petaflops, he says. 鈥淚n fact, the U.S. has 4,400 petaflops. And they are about to add another 2,000 petaflops to that.鈥
Canada lags even non-superpowers. Brazil has 53 petaflops, Poland 65, Italy 330, and Japan 672.
鈥淲e鈥檙e behind by orders of magnitude,鈥 he explains.
It鈥檚 one of the reasons he returned home to Canada 鈥 and to Queen鈥檚.
鈥淚 wish Canada was in better shape in the supercomputer world,鈥 he says.
鈥淥ne of the big challenges is that worldwide there鈥檚 just not a lot of supercomputer experts. We just don鈥檛 have a lot of experience in Canada building those really, really big systems. There are no dedicated institutions in Canada to building supercomputers and keeping us on the forefront,鈥 he says. 鈥淥ther countries have built up national laboratories that are constantly concentrating on questions like, 鈥楬ow do we build the next one? What are the bottlenecks we need to resolve?鈥欌
Dr. Grant hopes he can bring some of that expertise to Canada.
鈥淭hat was really my driving motivation 鈥 to provide opportunity for Canadians to enter the industry and to provide opportunities for Canada as a nation to be able to capitalize on supercomputing expertise to drive the economy, to solve productivity problems here in Canada, and just make us the world leader that we should be in this space,鈥 he says.
Why Queen鈥檚?
鈥淚 could have gone anywhere, but Queen鈥檚 has some advantages in this space. Queen鈥檚 has some of the fastest-growing AI research programs in Canada. It has accessible land. It has plenty of water. It has capacity for power. And it鈥檚 nice and conveniently located between Toronto, Montreal, and Ottawa.鈥
The location is important. With Canada鈥檚 limited supercomputing infrastructure, companies and institutions that need the power of a supercomputer might find they need to share that resource. Additionally, some supercomputing tasks are highly sensitive. Users may need to be using the computer directly, not logging in from elsewhere.
鈥淚f you鈥檙e training AI to consult with wealth clients at a bank, do you really want that information sitting on the internet? What about personal health data? Probably not.鈥
AI also plays a key role in national defence and signals intelligence and Queen鈥檚 is close to the Royal Military College, the Canadian Defence Academy, and other military institutions.
鈥淪uppose in the future another nation develops antagonistic AI that attacks all of our infrastructure. How do you defend that? You need defensive AI. How do you build that? You need a supercomputer,鈥 Dr. Grant says. 鈥淚n the future, battle can be these AI systems in cyberspace battling each other, trying to find ways to break through.鈥
Beyond that, Queen鈥檚 and Kingston are just wonderful places to live and work, he says.
鈥淚 think it鈥檚 really exciting to be back at Queen鈥檚 and to be contributing. I regard it not just as fun and a great place to be, but coming back to serve my own nation after working in the U.S. for so long. I鈥檓 glad to be back.鈥
Supercomputing 101
The adage 鈥渕any hands make light work鈥 is a pretty good analogy for the modern supercomputer, which really isn鈥檛 a single computer at all. The 鈥渉ands鈥 in this case are a network of thousands, even tens of thousands, of high-end processors working in unison on the same problem.
That problem could be finding new drug treatments, analyzing experimental results in subatomic research, or the study of astrophysics to help us understand our universe. In Japan, supercomputers are improving predictions of earthquakes and tsunamis. Environment Canada harnesses supercomputing power to improve its daily weather forecast.
Supercomputers gave us the two-metre rule used during the COVID-19 pandemic, while the explosion in the use of artificial intelligence wouldn鈥檛 be possible without the staggering power of supercomputers.
The largest supercomputer in the world covers about 650 square metres and would fit comfortably into the floorspace of Grant Hall. But it鈥檚 not their size that makes supercomputers 鈥渟uper鈥 鈥 it鈥檚 their speed.
鈥淲e judge these things in what we call 鈥榝lops,鈥 floating-point operations per second,鈥 Dr. Ryan Grant says. 鈥淏asically, it鈥檚 just a calculation. Can I take two decimal numbers and multiply them together? How many of those can I do in a second?鈥
In the realm of supercomputing, that number is astounding. 鈥淧eople understand the terms 鈥榤ega鈥 and 鈥榞iga鈥 because that鈥檚 how you measure the memory in your own computer,鈥 Dr. Grant says. But those measurements are inadequate when talking about the power of a supercomputer.
鈥淎 petaflop is a million billion operations per second. And an exaflop is a billion billion operations per second,鈥 he says. 鈥淭he world鈥檚 fastest supercomputer has just broken the exaflop barrier.鈥
The secret to supercomputing lies in how to make those individual processors communicate with each other. For that you need maximum bandwidth 鈥 the amount of information that can be carried 鈥 and minimal latency 鈥 the speed at which the information travels through the network.
Imagine a factory of 30,000 workers co-operating on the same task.
鈥淚t鈥檚 the co-ordination to get all those people moving together that鈥檚 hard,鈥 Dr. Grant says. 鈥淲hen we talk about communicating between systems, we鈥檙e talking in the hundreds of nanoseconds.鈥
Designing the software that controls those networks is Dr. Grant鈥檚 specialty.
鈥淭hat鈥檚 my bag of tricks. I do new network designs and I solve problems,鈥 he says. 鈥淚鈥檓 very proud to say that one of the network specifications I worked on 鈥 the software interface design for networks that helps dictate how the hardware is designed 鈥 that's in every single exascale supercomputer in the world.鈥