Challenges

Some of the coolest people at the Arctic Region Supercomputing Center during the summer are the latest crop of computational science interns, who converge at ARSC from various corners of the country. They are young, ambitious, energetic, stylish, sharp-witted, great at making carpool playlists and terrible at erecting tents. Hanging around with the ARSC interns is like experiencing a living, breathing, iPod-carrying version of the World Wide Web. In the course of any given half hour, they might remark that Google can now index Flash files, debate Tom Brokaw’s replacement on Meet the Press, recite the monologues of Eddie Izzard, and wonder what a rainbow would look like in a four-dimensional world. Being supercomputing interns, they know things like how to program simple search engines and how a flow model works. They like basketball, knitting, programming in C++ and Python, playing Spades, hiking, OAR’s live album, elegant equations, NASA, Thai food and horchata, but only if it’s from Mexico. They think the physicist Richard Feynman is “frickin’ Don Juan cool!” and find the views in Denali National Park “really awesome!” and “totally beautiful!” Like ARSC’s supercomputers and like the midnight sun, they never seem to rest.

It’s Wednesday morning and intern Ben Sperison is still up discussing politics with his roommate when Darren Lamison-White leaves the University of Alaska Fairbanks dorms for a three-mile run. He wears a “Midnight Sun Run” T-shirt and Billabong running shorts that extend past his knees. During the school year, Darren is a computer science major at Bowie State University in Maryland. He is one of 11 undergraduate interns in ARSC’s Alaska Research Summer Challenge, and joins four graduate student interns, three military cadets and visiting faculty at ARSC for hands-on, research-driven, intensive training in advanced high performance computational science.

Each undergraduate in the Alaska Research Summer Challenge has an internship project. As far as work goes, the project is what the summer internship is all about.

This is how the projects took shape: At the start of summer, Greg Newby, ARSC’s chief scientist and internship coordinator, paired the interns with mentors from ARSC staff and UAF faculty, who are working on HPC-related research projects of their own—anything from ice sheet modeling to weather forecasting to software and hardware design. Each intern took on a sub-project of his or her mentor’s larger, often ongoing research project. This generally works out great for the mentors, since it furthers their research, and for the interns, since they get to do original work. “They’re never solving problems that are solved problems,” Newby says. “There’s always an aspect of discovery of things that are not previously known.”

The thing that’s not previously known about Darren’s project is how well matrix arithmetic works for finding information in a very large collection of data—something that search engines like Google must be able to do in order to list web pages relevant to a query for, say, “horchata” or “Eddie Izzard monologues.”

Most of the time, Darren writes code on his laptop in his dorm room. He chose to write in the language C++ because he is familiar with it, even though the HPC specialist who’s been helping him would rather he write in C. “The guy’s like anti-C++. We’re always jokin’ around how he likes C, I like C++.” Darren says he also could have used Matlab to get the same results, which would have been a whole lot easier, but he wanted to create his own application. “Why? For the experience. It really helped me understand what I’m really doing. ’Cause if I just did it in Matlab, I could seriously go the whole summer spitting out results and not really knowing how I got to that point.”

When Darren has to test his program, he goes to ARSC to run it on Midnight, a 2,312-processor Sun Opteron supercomputer. After he showers and eats a bowl of Cheerios, he walks the quarter mile across campus to the lower level of ARSC, which Darren says “you’re not allowed to call the basement.” Besides lacking windows, the lower level ARSC offices look nothing like a basement anyway. The room is full of light and divided into cubicles, which the interns share. A sign on Darren’s cubicle says:

A few cubicles down the hall, someone has drawn the following picture on a white board:

On this particular day, Darren is pleased with his project because he just figured out “like yesterday” how to make his program multiply matrices. “It’s a great sense of accomplishment when you finally get what you’re working on,” he says.

While Darren touches things up, Princess Trillo turns on a monitor in a cubicle around the corner. She says she usually doesn’t like to work at ARSC because “it’s too loud in here,” although at the moment the only sounds are the whirring of processors and the clicking of keyboards. Princess is a computer science major at the University of Texas at El Paso. She grew up across the border in Juarez, Mexico. She wears black slacks and Converse shoes and a bracelet made of Starburst wrappers. When she types, she has perfect posture, which makes her look very professional. When she speaks, she often tucks her long dark-brown hair behind her ears, which makes her seem a bit shy.

“I didn’t grow up around computers,” she says. “My dad, up until a year ago, didn’t even know how to turn on a computer.” She was “very excited,” she says, to learn she’d been accepted to ARSC’s internship program because she wanted to come here last year but didn’t apply. “I guess I thought I didn’t have enough experience and I didn’t know enough to come here. And now that I’m here, I realize I still don’t know enough,” she laughs.

Princess works with ARSC HPC Specialist Anton Kulchitsky, whom she says is “super smart.” Kulchitsky is developing a set of programs called the Solar Wind Delay Project. The programs are tools to help researchers create and test models for understanding how solar wind reacts with the Earth’s magnetic field—a phenomenon that produces the aurora borealis and can knock out power grids and disrupt satellite communications. Princess’s job is to modify the programs, run them on Midnight and graph the output data.

It’s now lunchtime, and Ben Sperison is buying a Beef and Cheddar BBQ sandwich at the Pizza Piazza on campus. Ben is a junior at New Mexico State University. He was double-majoring in computer science and physics until he added economics as a third major. He wears a Casio sports watch, jeans and running shoes. His eyes are puffy. He says he doesn’t usually stay up until 7:30 in the morning but admits that he sometimes keeps such odd hours at work that UAF Security once put a tail on him.

Ben works at the Department of Mathematics and Statistics a few offices down the hall from his mentor, Ed Bueler. Bueler is developing an ice sheet model called PISM, which simulates how ice flows atop Greenland and Antarctica. Ben spent the first few weeks running PISM and making visualizations of Greenland’s ice sheet from the results. He finished with that much quicker than Bueler expected and thought he might like to learn the mathematics behind ice sheet modeling. “[Ben] could see what the Greenland simulation looked like and wasn’t that interested in spending more time making it look better or doing harder, higher resolution runs and post-processing the results,” Bueler said. “Instead he decided he wanted to know what was going on inside the black box—how the internals of this program work. The majority of the summer he spent learning what we teach in a graduate course.”

Ben finishes his sandwich and says he had better get back to work. He has a lot of reading to do.

Bethany Waldmann is also just getting back to work. She usually works in her dorm room because she says the chairs at ARSC make her slouch. She attends Messiah College in Pennsylvania and, according to the colored bands on her wrists, supports the college’s sports team, the Messiah Falcons, and research to find treatments for myeloma. She used to be a math education major but recently switched to computer science because she likes programming “way more than statistics.” She says: “As part of the math program, you’re required to take Programming 2 or Stat 2, and I was like, ‘Totally Programming 2, are you kidding me?’”

Bethany’s internship project requires a great deal of programming. She is working with Greg Newby and Don Morton to write scripts that will automatically convert illegible data sets into colorful web graphics for ARSC’s smoke forecasting website (http://smoke.arsc.edu/). As Bethany explains it, ARSC put together a team of scientists a couple years ago to develop a smoke-forecasting model for the state of Alaska. Less than two weeks before she arrived in Fairbanks, the team had made enough progress to run the model daily on Midnight. The model spits out forecasts at the end of each run in the form of a huge data file. Made up mostly of numbers, the file is over a million lines long and not particularly pleasant to read. Bethany’s scripts take this file data, plot it onto a map, and upload the images onto ARSC’s webpage.

Bethany says she didn’t know what she expected from the ARSC internship, but she does know she didn’t expect the interns to have such varying levels of experience. “There’s a couple interns who just finished their freshman year, and like Cory over there,” she whispers, “he interned at NASA…”

It’s true that Cory Simon once interned at NASA. It’s also true that Cory has gone deep-sea fishing, scuba diving and sky diving…twice. It’s probably safe to say there’s not a lot that Cory hasn’t done, and nearly everything he does he describes as “the most amazing thing.” Yesterday, he took the Graduate Record Examination, and while he doesn’t say it was amazing, he does say he wants to celebrate.

At the Marlin, where giant plastic fish cover the walls and local beer is served in mason jars, Cory talks about his “totally amazing” experiences at ARSC this summer. For example: “The hike on Angel Rocks [part of the one-million-acre White Mountains National Recreation Area just north of Fairbanks] was really cool and I thought that was the most amazing thing, being able to see the views. And there were F-16s training in the mountains and they were flying around and that was so amazing.”

Cory is a senior at Iowa State University and a computer engineering major. Like Princess, he’s working with Anton Kulchitsky on the Solar Wind Delay Project. While Princess post-processes data, Cory pre-processes. Cory explains it like this: There are currently two satellites that sense the solar wind—one closer to the sun, the other closer to Earth. Ideally, solar wind modelers can take the data from the satellite nearest the sun, plug it into their models, and predict the behavior of the solar wind. As the solar wind travels toward the Earth, modelers then take data from the second satellite to test the accuracy of their predictions. The problem is that data from these satellites is stored in very different formats, making it difficult to access. So Cory is writing code that modelers can use to easily access data from both satellites without having to learn anything about its format.

Next door to the Marlin is the Pad Thai restaurant, which the interns frequent because they can walk there from campus. Mike Jacobi shows up right at 7 p.m., even though he was just starting to make a lot of progress on his project. “I’m moving now,” he says. “Before I was stuck, I hit a wall.”

Being a sophomore at the University of Montana, Mike is somewhat new to programming. Still, he figured out how to run the Weather Research and Forecast (WRF) model for historical weather events, compare the forecasts to observations, create visualizations of the data, and organize the results in one place: http://weather.arsc.edu/StudentWork/jacobi/

Particularly, Mike has been looking at weather inversions, which happen when cool air pools below warm air and which WRF is notoriously poor at forecasting. All of this helps his mentor, Don Morton—who runs WRF on Midnight to make daily weather forecasts—measure and improve the accuracy of the WRF model.

Mike says he’s learned more this summer than he did all last year at school. Sometimes when Mike isn’t learning and is instead hitting a wall, he talks to Cory. “I think I’m going to remember Cory most out of everyone. I want to be where he is when I’m his age. I admire his work ethic. And he’s a great leader. He helped me with so much this summer, and more than just work. He tricked me into volunteering for leading the Denali trip.” (An outdoors adventure weekend for the interns in Denali National Park and Preserve.)

As Mike heads back to the dorms, Cory walks back up the hill to ARSC to get in a couple more hours of work on his project. It’s now 8:30 p.m. and the sun is still high in the sky. Cory turns on his iPod. Behind him, Midnight whirs nonstop in the lower level of a handsome building overlooking the UAF campus and the Tanana Valley below.