Science 1663

Information Science and Technology Revolution

From Terry Wallace

Computer simulation of complex systems is a hallmark capability at Los Alamos National Laboratory.

It began with John von Neumann, who joined the Manhattan Project in 1943. His very difficult calculations concerning nuclear explosions made him realize that machines could be used to solve the numerical formulations of partial differential equations. This ultimately led to the Laboratory's first supercomputer: the Mathematical Analyzer, Numerical Integrator and Computer, affectionately known as MANIAC.

Ever since then, Los Alamos has pushed the frontiers in computers, building and operating the world's fastest machines, including the Cray-1 in the 1970s, Thinking Machines' CM-5 in the 1980s, and later this year, in collaboration with IBM, the first petaflop computer (a million billion operations per second)—Roadrunner.

Roadrunner increases performance through a dramatically new strategy—mixing different types of processors that can be optimized for different types of calculations. This strategy holds the promise of computers hundreds of times more powerful than Roadrunner within the next decade.

Hardware is only a small part of simulating complex systems. Stanislaw Ulam, another Manhattan Project member, had the idea of evaluating complex processes statistically by what is now called the Monte Carlo method. Legend has it that while playing solitaire, Ulam thought about calculating the probability of winning, but realized that playing many games and counting the number of successful plays would give the answer more easily. He then realized that the new fast computers could use that strategy for problems of neutron diffusion and other processes described by certain differential equations, provided those processes could be translated into an equivalent form composed of a succession of random operations.

Ulam's idea gave birth to one of the Laboratory's most-powerful computer simulation tools, the Monte Carlo code MCNP.

Today, Los Alamos has world-class capability in simulation science and a focus on information science and technology (IST), reflecting the growing need to solve data, information, and knowledge issues in 21st-century science. From satellite-based sensors to medical imaging devices and intelligence databases, automated collection of terascale data is becoming standard. The generation of petascale simulation data is close at hand.

This edition of 1663 is a window into the IST revolution to come.