Communications of the ACM
One-Hundred Supercomputers Later, Los Alamos Still Pushes the Hpc Curve
Installed in 1957, the MANIAC II's vacuum tubes were replaced by solid-state circuitry before the system was decommissioned.
Credit: Los Alamos National Laboratory
From the 1952 MANIAC to Bonanza deployed just this month, Los Alamos National Laboratory has deployed 100 supercomputers in the last 60 years — a showcase of high-performance computing history.
"The computing capability in our data centers in any given year dwarfs what was there 10 years before," says Randal Rheinheimer of the High Performance Computing division, "and Los Alamos has been on that curve for 60 years."
The Los Alamos computers deployed along the way include the MANIAC II, which started its nearly 20-year service life with over 5,000 vacuum tubes, all of which were replaced over time with circuit boards. Any computing device today would have to be purpose-built to compute as slowly as the MANIAC computers. But compared to a room full of people with mechanical calculators, those early computing tools were significant advances, Rheinheimer says. "It was the first and only triumph of serial over parallel computing."
Other deployed systems include Stretch, a technological stretch built in collaboration between Los Alamos and IBM, Serial Number 1 of the iconic Cray-1, and a Thinking Machines CM-5, with its lightning bolt footprint and fat-tree interconnect. "The fat-tree today seems an obvious topology, but we have detailed diagrams of it in our archive labeled 'Thinking Machines Proprietary,'" Rheinheimer says.
By 2008 there came Roadrunner, the world's first petaflop supercomputer, which provided a boost for unclassified science before being moved to the secure side to perform feats of computational heavy lifting for the nuclear stockpile stewardship program. A mere five years after installation, its powers were eclipsed by the next generation of big iron and its components were repurposed.
"Computing power for our Laboratory's national security mission is a huge part of our proud legacy, and it plays an integral role in our bold future," says Laboratory Director Charlie McMillan. "The continued innovation and enhancement of this capability touches areas from stockpile stewardship, to global security simulation, to biological modeling, and even safety basis calculations. I look forward to our next 100 supercomputers and their impact on the nation."
These days, the Lab has 13 supercomputers, or "high performance computing clusters," operating in both unclassified and classified environments. With names like Cielo, Lobo, Mustang, and Moonlight, the work performed by Roadrunner has been shifted to other machines. A time-lapse video of the installation of Cielo, a current stalwart of National Nuclear Security Administration capability computing, accompanies this article.
The computer review is part of Los Alamos' burgeoning HPC History Project, in partnership with the Charles Babbage Institute at the University of Minnesota. An infographic produced as part of that effort illustrates highlights of the lab's supercomputer deployments.
The project has been unearthing a wealth of information about supercomputers and the supercomputing industry, as viewed through the prism of 70 years of computing for science in the national interest at Los Alamos National Laboratory.
No entries found