The center's resources for interactive supercomputing and high-performance data analysis have been critical to R&D programs at Lincoln Laboratory.

The Lincoln Laboratory Supercomputing Center (LLSC) celebrated a significant anniversary this year, marking five years of the center's mission to enhance the computing power available to Lincoln Laboratory, MIT, and other researchers.

The LLSC was established on April 1, 2016, to build on the Laboratory's past groundbreaking work in computing. In the late 1940s and early 1950s, prior to the establishment of Lincoln Laboratory, MIT designed and built the Whirlwind 1 computer using vacuum tubes for the U.S. Navy. Whirlwind was unique for its ability to output information in real time and perform calculations in parallel. Later, in 1955, the first fully transistor-based computer TX-0 was designed by staff at Lincoln Laboratory, which had been established four years prior.

These and other early computing accomplishments eventually led to the establishment in 2003 of LLGrid, a Laboratory supercomputing system. In 2008, the Laboratory demonstrated the largest problem ever run on a computer using the TX-2500 supercomputer, a computer that was a part of LLGrid.

The LLSC was established with a focus on interactive supercomputing and high-performance data analysis. The staff who started the LLSC decided to model the center on the partnership model found at federal laboratories and universities around the world.

"In any given year, over half of the Laboratory staff are conducting supercomputing-enabled research. Our hallmark is being the world leader in interactive supercomputing," says Jeremy Kepner, who is a Lincoln Laboratory Fellow in the LLSC. "When we started, all processors had a single core, accelerators did not exist, and most of the work was signal processing in MATLAB. Today, all of our processors have at least 20 cores, we have hundreds of accelerators, and the work spans many domains using a wide range of programming environments."

The resources that LLSC provides have been critical to many award-winning technologies at the Laboratory. For example, a Laboratory-developed program called Large-scale Automated Vulnerability Addition (LAVA), which helps discover bugs in code, utilized the LLSC for a competition they hosted. The competition, called Rode0day, pitted programs based on LAVA against each other. LAVA won a 2020 R&D 100 Award.

"Using the Laboratory's supercomputing resources, we were able to run an experiment to test eight of the leading bug-finding systems," said Andrew Fasano, a researcher in the Cyber System Assessments Group who worked on LAVA. "By using a standard one-CPU system, it would have taken more than 83 years to collect all of the data."

Many of the staff who were there for the start of the LLSC are still at the Laboratory today.

"We continue to add new people to the group, but being able to take on this journey with this team has been particularly special," says Vijay Gadepally, a senior staff member in the LLSC.

The LLSC deployed TX-GAIA in 2019, which is currently rated by Top500 as the 57th fastest supercomputer in the world. It has served a critical role in AI research, including research conducted by the Department of the Air Forceā€“MIT AI Accelerator and research related to COVID-19.

In the future, the LLSC hopes to support even more research and collaborate with more researchers across the Laboratory and MIT.

"Collaboration with regional partners will be a key part of taking our next big step, and will allow us to do bigger science, engineering, and supercomputing together," says Kepner.

"Supercomputing plays a huge part in research going on across the Laboratory, and LLSC systems have provided the computational power for a number of Laboratory achievements," says Gadepally. "We've been working to ramp up the research output from the LLSC to become one of the nation's preeminent supercomputing research organizations, in addition to the cutting-edge infrastructure and systems we already provide."