Space weather and cloud computing, a new forecast for science

Doctoral student Jan Paral and professor Robert Rankin are using cloud computing to model conditions surrounding Mercury.

Sandy Robertson - 18 December 2012

A U of A space physics doctoral student and his PhD supervisor are stretching the capabilities of cloud computing-by roughly 154 million kilometres-to model the conditions surrounding Mercury. In the process, they and their colleagues may help fundamentally reshape the way scientists access, harness and share computing power.

Jan Paral was able to create detailed (and visually stunning) animated models of Mercury's magnetosphere and its interactions with the solar wind, by combining satellite data with observations from MESSENGER, NASA's spacecraft orbiting the planet. Paral's models incorporated figures and animations produced by Chris Want from the University of Alberta's Academic Information & Communications Technology division.

Paral's code requires a bit more computing power than the average PhD student can access. Fortunately, he found it-in the form of BlueGene, IBM's massive supercomputer. Paral spent time at IBM's T.J. Watson Labs in New York, porting his code to BlueGene, and optimizing and debugging it. Now, when he needs a supercomputer, he can access BlueGene's immense processing power from his desk at the U of A.

"The computer itself fills the entire room," Paral marvels. "There are several racks, just like big bookshelves, and each of these shelves has roughly four thousand processors. I generally use one or two of these shelves for a couple of days."

Paral doesn't exactly spend that time playing Tetris. "One of my bigger simulations needs more than three billion particles pushed step by step, followed in space and time, to get the kind of result you see on the animation," he explains. "You need an enormous amount of resources to do that, for several days."
As Paral's PhD supervisor, space physics professor Robert Rankin (Physics) gets a kick out of seeing a protégé working with one of the most powerful supercomputers on the planet. "It's quite notable that IBM has given Jan this special access. He's one of very few that I'm aware of who have that access. They don't let just anyone walk in the door."

Paral's work is the boldest example so far of the power of CESWP-Cloud-Enabled Space Weather Modeling and Data Assimilation Platform. The U of A-driven project uses cloud computing technologies to put powerful new tools into the hands of researchers. In the CESWP environment, student scientists can access preconfigured space weather simulation tools, grab powerful computing resources when they need them, and build on each other's results.

CESWP grew out of Rankin's desire to develop a more productive and powerful way for his students to work. Until now, he explains, space physics grad students have been forced to spend enormous amounts of time just configuring their own computers and servers and learning how to use them. "Students can get caught up in doing things that have nothing to do with their research." Then, when one graduate student leaves the program, the next must begin essentially from scratch. "That computer is likely scrubbed for another student," Rankin says. "You don't have a legacy of what they've done."

With CESWP, future grad students should be able to simply sit down, roll up their sleeves, and get to work. "I can sit at any computer, launch on our cloud-a virtual computer-and then use that just like my real computer," explains Rankin. "What I connect into is preconfigured with everything a person would need, or we can put it there if we don't have it. You instantly have access to a common set of tools and infrastructure you need to do your research."

The system also allows students to tap into the work of their colleagues, building collective expertise layer upon layer, Rankin says. "It serves as a backlog of your work, and at the same time it works as a gateway to models that someone wrote before you. You just click and you've got the results, instead of spending time rewriting models that were already there. So, if someone wanted to use Jan's code, he points them to the virtual environment and they're off and running."

CESWP was developed by software experts from the U of A's AICT unit and from Cybera, the non-profit organization that oversees e-infrastructure in Alberta. Major funding was provided by CANARIE.

Barton Satchwill (BSc '97), a senior developer at Cybera, relished the opportunity to help Rankin's vision take tangible shape. "Robert was smart enough to recognize the potential of this way of doing things," he says. "It was enormously challenging, because there were no recipes that we could follow. There were no guidebooks for this sort of thing. It was new territory."

Satchwill predicts CESWP will prove to be an academic game-changer. "It's certainly going to be the way that almost everything is done in the future."