Scientists, IT community await exascale computers

Computerworld - The race is on to develop a new generation of far more powerful supercomputers that could help solve some of the world's most vexing problems.

Exascale supercomputers, expected to appear by 2018, could, for example, play a significant role in efforts to combat climate change or develop ultra-long-life batteries for powering automobiles. Scientists also expect exascale systems to help them come up with processes for creating biofuels from weeds rather than corn.

Much of the work at various national labs to design and develop the new systems is funded by corporations that hope their IT operations can take advantage of the new technologies.

For example, the next generation of supercomputers could be used to solve big programming problems and allow for the development of a new generation of scientific and business applications.

The need for exascale systems, and the difficulties developers face in trying to boost hardware performance without soaking up excessive megawatts of power, was widely discussed among many of the estimated 11,000 people who gathered last month in Portland, Ore., for the SC09 supercomputing conference.

"There are serious exascale-class problems that just cannot be solved in any reasonable amount of time with the computers that we have today," said Buddy Bland, project director at the Oak Ridge Leadership Computing Facility in Oak Ridge, Tenn.

The world's fastest supercomputer today, a Cray XT5 system at Oak Ridge National Laboratory that's known as Jaguar, has a peak performance of 2.3 petaflops. A petaflop is a quadrillion, or 1,000 trillion, sustained floating-point operations per second.

The total capacity of the latest Top500 list of the most powerful supercomputers, released at SC09, was 27.6 petaflops, up from 22.6 petaflops in the previous list, released in June.

One exaflop is 1,000 times faster than a petaflop -- performing 1 quintillion, or 1 million trillion calculations per second. "We think exascale is a 100 million-core kind of enterprise," said Dave Turek, vice president of deep computing at IBM.

In mid-2008, IBM's Roadrunner supercomputer -- a hybrid system that runs both AMD's Opteron processors and Cell chips designed by IBM, Toshiba Corp. and Sony Corp. -- was the first to achieve petaflop speeds. Now the U.S. Department of Energy has started making plans to build an exascale system that's 1,000 times more powerful than Jaguar.

These future systems must use less memory per core and more memory bandwidth. Systems running 100 million cores will face continuous core failures, and the tools for dealing with them will have to be rethought "in a dramatic kind of way," said Turek.

Addison Snell, CEO of InterSect360 Research, expects general-purpose exascale systems to come out of the supercomputer research efforts, though he predicts that "special-purpose [systems] will probably come first."

Stephen Lawson of the IDG News Service contributed to this story.

Read more about Hardware in Computerworld's Hardware Topic Center.