Livermore computer scientist and former Lawrence Scholar Olga Pearce has spent the past few years working with her Center for Applied Scientific Computing (CASC) colleagues on the problem of load imbalance in high performance computing applications. Load balancing is the practice of distributing work on a parallel computing system in a resource-efficient manner.
When some processors are idle while others are working, the load is imbalanced. Application programmers try to strategically assign work when writing code, of course, but when simulating complex physical phenomena, such as a supernova or a nuclear explosion, it is not always possible to predict which portions of the simulation will require the most processing effort to calculate. This is where researchers like Olga can help.
Olga specializes in designing and implementing strategies for detecting and dynamically correcting load imbalances while the application is running. The method she uses is scalable and works for many different physics codes.
Says Olga, “I’m helping real applications run faster, but I also get to learn about physics and other areas of science at the same time. From my perspective, that’s the cool thing about my job. This is why I could imagine myself spending my career here—I doubt I’ll run out of things to learn.”
It took courage and perseverance for Olga to pursue her chosen career path. As a young adult, she left her family in Lviv, Ukraine, and moved to the United States to find work. “Change is easier at that age,” she notes. “You adjust to new things well.”
Olga was able to enroll in a nearby college soon after. She started by majoring in business for practicality reasons, but eventually switched to her two favorite subjects, math and computer science. “It’s a privilege to get to do what you want to do, career-wise,” says Olga. When she decided to continue on to graduate school, she focused on computer science because it seemed to offer a wider array of research opportunities.
Olga met her husband, also a CASC computer scientist, when they were both in graduate school. When deciding where to start their careers, the San Francisco Bay Area, and LLNL in particular, seemed like a promising spot for two computer science researchers. Both participated in summer internship programs and found they liked it here. Lawrence Scholar Program tenure only solidified their hopes for interesting research careers at LLNL.
Now, having earned her Ph.D. from Texas A&M University and accepted a staff scientist position at Livermore, Olga is continuing her research into parallel application performance. “Load balancing isn’t a solved problem and won’t be for a while,” she says. “Also, the problem changes as machines change. As we get newer and larger computers, new performance challenges come up. I don’t foresee running out of interesting problems to solve.”