George Rebane
As technology accelerates, one of the measurables is how often we have to upgrade our notion of supercomputer and how often supercomputer users must buy new ones. The Department of Defense uses a lot of supercomputers and the pace of development has become so hectic that it must now upgrade every two to three years. More here.
The corollary to all this new supercomputing hardware is supercomputing software which is having a hard time keeping up. To get the required performance out of new supercomputer architectures (how the components are hooked up and communicate with each other), new kinds of programming approaches are needed. There is a lot of parallel work going on here too. We will all be hearing a lot about “petascale computing” which requires engineers to rethink whole new approaches to how problems are formulated and numbers are crunched. Supercomputers gain much of their speed through parallel processing where thousands of computers work concurrently on parts of the problem passing data from one to another in a whirlwind.
Petascale computing refers to the number of basic operations – add, subtract, compare – a computer can do in a second. The box on you desk or laptop is pushing gigascale or gigaflops (floating point operations per second), which is in the order of a billion or 1,000,000,000. Petascale is a million times faster, or a 1,000,000,000,000,000 flops. At these speeds a whole new world of possibilities opens up.
But to take advantage of petascale requires education now hard to come by in regular computer science curricula across the country. The National Center for Supercomputing Applications has been working on this problem with their Blue Waters virtual school program where they run “site where students can learn about petascale computing for science and engineering.” Professor Sharon Glotzer (University of Michigan) states,
"Many aspects of the nuts and bolts of computational science ... fall between the cracks, and as a result, it is not easy for today's students to learn all they need to know to become tomorrow's innovators in high-performance scientific computing," She says the mission of the virtual school is to fill in the gaps in students' knowledge, particularly in petascale computing. "For many of our most important scientific applications, petascale computing will force us to rethink how we structure our codes to take full advantage of the architecture of these new machines," Glotzer says.
More here. Now you can, perhaps, get a feel that there is going to be an historic surge in science instead of what some people are advertizing as the “end of science.” And where will our Jane or Johnny fit into a world with petascale computing?
Comments