Jimena González Lozano and Aashish Tripathee are 2023’s recipients of the award for their research advancements with strategic use of high-throughput computing (HTC).

OSG leadership created the OSG David Swanson Award in memoriam of Swanson, who championed throughout his life for both the success of his students and the expansion of OSG and research computing. David Swanson, who passed away unexpectedly in 2019, was a computer science and engineering research professor at the University of Nebraska-Lincoln. The award reflects Swanson’s and OSG School’s emphasis on helping people develop their skills in technology and advancing science with large-scale computing, OSG research facilitation lead Christina Koch says.

Researchers — like Jimena González Lozano and Aashish Tripathee who sought the OSG School’s high-throughput computing (HTC) resources to solve complex computational challenges, and in turn, were able to evolve their research projects — have been honored with the award since its establishment in 2019. González is a Department of Physics observational cosmology Ph.D. student at the University of Wisconsin-Madison, and Tripathee is a University of Michigan Physics post-doctoral research fellow.

Awardees are provided the opportunity to share their research at the OSG All-Hands Meeting, which is part of the annual 2023 Throughput Computing (HTC23) conference, held in Madison, Wisconsin. “To have it in the context of recognizing a wonderful person like David is really meaningful. It’s like ‘Oh yes, this is why we’re doing what we’re doing,’ and it’s rewarding,” Koch reflects.

As a David Swanson awardee, it’s an honor to be an example of how HTC and the OSG School transformed her research, González elaborates. “I couldn’t even explore new ideas [because it could take weeks to run one simulation], and it was around that time that I was reading all my emails carefully, and I saw the OSG User School [application] announcement,” González remembers. “They did a really good job at describing what you would learn and what high-throughput computing is. From that description, I thought that it was perfect for me. I applied, and then during the summer of 2021, I learned how to implement it, and it was very quick. After the first day, I already knew how to submit a job.”

Gonàzlez’s research on strong gravitational lenses in the dark energy survey implements HTC and machine learning. Strong gravitational lenses can image stars from which González can extract the position of the source and the magnification between the images. From the images, González creates thousands of simulations composed of millions of images while constraining the quality of the images. Because of the volume of simulations she needs to train, González could be left waiting for up to weeks using machine learning — and the tighter constraints, the greater the waiting time. This put its own constraints on which properties she could experiment with. Some ideas, González says, were impossible to do because she couldn’t do them quickly. Implementing HTC shortened the waiting time from days to hours.

The OSG school also impacted other areas of González’s research, including training the machine and performing a complete search — each was reduced from long wait times spanning days to years to much more manageable wait times of as little as three hours.

Tripathee uses HTC for solving a big data challenge too. For one project on continuous gravitational waves, the data he collected spans a year and the entire sky, as well as the polarization over 24 times, resulting in 80 quadrillion templates. The solution, Tripathee said at HTC23, is looking at 500 billion templates per job. The answer for computing templates at a magnitude of a quadrillion is to use HTC, which helps with efficiency when running the numbers and makes the project possible. Without HTC, Tripathee’s jobs would’ve taken on average more than 10 hours for some or more than 24 hours for others. Through the OSG, Tripathee uses 22 million core hours, 1.4 million hours per month, and 47,000 hours per day.

Tripathee’s mentor and OSG Deputy Executive Director Tim Cartwright encouraged Tripathee to self-nominate for the award. Upon learning he was chosen to receive the award, “It felt like a nice validation and a recognition of having used the [OSG] to perform research,” Tripathee says about receiving the award. Attending HTC23 event in Madison to receive the award was rewarding. “I also got to meet a lot of people… like the OSG faculty, Tim Cartwright in particular, and Christina [Koch]. There was a really nice opportunity and an honor to come to Madison, attend the event, and receive the award but also meet [David Swanson’s widow, Ronda].”

At HTC23 Ronda Swanson said she hopes David’s legacy will live on in science. Ronda Swanson is OSG’s self-described “biggest fangirl” and has continued her relationship with the OSG as an advocate for HTC since David’s death, Cartwright says.

Annually, a committee chooses one or more former students from the OSG School according to the student’s advancements and research achievements with distributed high-throughput computing (dHTC). The OSG School teaches students how to harness HTC resources for their data collection and research needs. Koch, who served on the selection committee for the award, explains the committee looks for students who have taken what they learned at the OSG School and achieved great things with it, like tackling a research problem or writing workflows from scratch after coming in with little to no experience. Cartwright says committee members also look for applicants who can give back to the community. Both González and Tripathee embody what the selection committee looks for, Koch explains.

“What Jimena learned [from the OSG School] really helped her solve a problem that she wouldn’t have been able to solve before. Aashish is tackling both a niche field of research with these resources and also has been testing new features for us or letting us know when things aren’t working and has had this ongoing relationship with us.”

González will continue to use HTC to model the mass distribution of each galaxy that produces a gravitational lens. People previously performed the computing for these models by hand, but as the data accumulates, it becomes less feasible for humans to do this computing. To remedy this, González will use machine learning to do the modeling because it requires a great deal of computational power.

Tripathee plans to continue using the OSG’s resources on new data and to conduct deeper searches more quickly and efficiently. “With OSG, we didn’t have to fight and struggle for resources. Having this access to these extra resources allowed us to do searches that are more computationally costly and sensitive,” Tripathee says. “If I had never heard of OSG, I would have probably still performed similar searches but not to this depth or sensitivity because the number of features that I would have had access to would have been more limited.”

“Once I was at [HTC23], I understood what impact he [David Swanson] had on people, and not only in developing OSG, which was huge,” Gonzàlez notes. “It was shocking, that impact, but it was so very interesting to see people talking about him because it seemed like he was also a really good human being, a really good mentor, and really liked helping people and supporting people.”