Supercomputer Provides Best Image Yet of Jet Engine in Action
Engineers have harnessed more than a million computer cores at once to run a simulation of a supersonic jet engine.
The project demonstrated that it's possible to model fluid dynamics on more than a million cores, according to Stanford University's engineering department, which ran the experiment in the 16-petaflop, 1.6 million-core Sequoia supercomputer in northern California. This may be one of the last publicly shared experiments Sequoia will run. The supercomputer will transition to performing classified work this March, according to the Lawrence Livermore National Laboratory, which houses and maintains Sequoia.
Jet-engine simulations help engineers develop quieter designs for planes, a boon for both workers on the tarmac and people who live near airports. The computer models allow researchers to see and measure what's going on in an environment that's often too hot and turbulent for direct measurements. Such models are so complex that only modern supercomputers are able to run them accurately, Parviz Moin, the director of Stanford's Center for Turbulence Research, said in a statement.
The Sequoia experiment is a big step forward for modeling at the center. "These runs represent at least an order-of-magnitude increase in computational power over the largest simulations performed at the Center for Turbulence Research previously," said Joseph Nichols, a center researcher who worked on the simulation. "The implications for predictive science are mind-boggling."
Sequoia previously ran a model named Cardioid, which simulated the communication between cells in the human heart. Sequoia has also run a simulation of the known universe, called Hardware/Hybrid Accelerated Cosmology Codes, or HACC for short.
In March, Sequoia will be dedicated to testing the United States' nuclear deterrent system. Among other things, the computer will be able to show what happens at an atomic level when a nuclear weapon goes off.