Livermore’s core mission of stockpile stewardship is built upon a modern foundation of carefully constructed simulation codes running on powerful computers. But as new supercomputers come online with new capabilities and architectures, Computing relies on engineers like Stephanie Brink to keep the legacy codes running smoothly.
“You’re only as fast as your slowest processor or your slowest function,” says Brink, who works in the Center for Applied Scientific Computing (CASC). By analyzing a legacy code’s performance, Brink and her team can reduce the amount of time it takes to run and allow for more critical science to be accomplished.
“I want to understand what function call is consuming the most amount of time,” she says. If one function is taking up 80 percent of the code’s runtime, Brink breaks it down into its component subfunctions and works with application developers to pinpoint the snag. On a formidable machine like Sierra, this approach can free up thousands of nodes, save on power consumption, and improve efficiency.
As a doctoral student at the University of Oregon, Brink worked on innovative ways to run applications on a computer where power is at a premium. But she never would have arrived in Eugene had it not been for LLNL.
“I was always thinking I’d go straight from undergrad to industry,” she remembers. “I hadn’t thought of other paths. It wasn’t until I came to Livermore and saw these other options that I opened my eyes toward going to graduate school.”
Her undergraduate computer engineering degree at the University of the Pacific required a summer internship. Though she held interviews with firms like NVIDIA, Intel, and AMD, she had heard so many great things about the Laboratory from an academic perspective that she landed in the Tri-Valley. Brink didn’t have much exposure to high performance computing at Pacific, so LLNL served as an introduction. While an intern with Computing during the summer of 2012, Brink attended SC, the International Conference for High Performance Computing, Networking, Storage and Analysis. For 48 hours as part of the Student Cluster Competition, Brink and her fellow students worked together to run a series of applications on a small supercomputer of their own. Her Livermore network helped provide her an opportunity to meet with potential grad school advisors.
“There was a huge, wide community of people doing this research; more than I ever could have imagined,” she says. “Going to grad school as the next step just made sense at that point.”
While working on her PhD, her connections led her back to the Lab in 2016 on a Livermore Graduate Scholar Fellowship. Her advisor at Oregon, professor Hank Childs, had spent 10 years working at LLNL and several years at Lawrence Berkeley National Laboratory. He advised her on her dissertation: “Optimizing Visualization Performance on Power-Constrained Supercomputers.” When the time came to move on with degree in hand, the path forward was both familiar and exciting.
“Going back to the Lab had always been a thought,” says Brink. “More than a thought. It had always been in the back of my mind that [LLNL] is somewhere I can see myself. Looking for jobs out of grad school, that was my bar: the experience I had at Livermore, and the flexibility I had.”
With CASC, Brink is afforded the ability to solve sophisticated computing problems that help her fellow scientists solve their own research problems. It’s a rewarding way for her to advance a broad range of sciences, whether on the open-source side or in the classified space.
“Part of our mission is always being one step ahead of things,” she says. “We like to say we’re ‘doing more science,’ and if we can improve the amount of time it takes for us to do more science, we can make better advancements in our fields and the mission overall.”
And she is happy to give back to the organizations that helped her along the way. For the fully virtual SC20 conference, Brink served on the committee for the Student Cluster Competition that helped open her eyes to the vast communities in HPC and at Lawrence Livermore. The variety in Computing and the Lab’s impact make for a compelling combination.
“People have been at the Lab for their entire careers,” she says, “but they’ve had three different careers at the Lab, transitioning between projects. I love the idea of having the flexibility to work in different roles. If I wanted to be a developer, I could do that; if I wanted to mentor students, I could do that.”