Publications
Attached Images
Peta Computing's Parallel Universe
by <!-- InstanceBeginEditable name="Feature2Author" -->Gordy Slack<!-- InstanceEndEditable -->
.................................................. <!-- InstanceBeginEditable name="Feature2Story" -->CITRIS researchers will soon have access to a new generation of high-performance supercomputers far more powerful than those available today. Known as petascale computers, these new machines will be capable of conducting 10^15 floating-point operations per second (petaflops). These parallel machines may employ more than a million processors and will be able to handle huge data sets. Until now, they have been mainly the domain of military and other national security applications. With the delivery of a new petascale computer to Lawrence Berkeley National Laboratory (LBNL), and with the possibility of a Berkeley team helping to host another one at Lawrence Livermore National Laboratory (LLNL), researchers working on climate analysis, genomics, environmental monitoring, protein analysis, earthquake, nanoscience, and other CITRIS-related fields will gain access to powerful new modeling tools within the next four years.
By employing a much higher resolution of analysis, seismologists here, for instance, will be able to do block-by-block modeling of earthquakes at different intensities, according to James Demmel, Professor of Mathematics and Computer Science at UC Berkeley and founding Chief Scientist at CITRIS.
James Demmel, CITRIS founding Chief Scientist.
"Until now, models have said, 'this huge area will vibrate about like this.' But that is not good enough to figure out which buildings need which kinds of retrofitting," says Demmel. "But with a petascale machine, you can refine the resolution of your simulations to determine which blocks and buildings are especially endangered and how best to retrofit them. It would enable a science-based approach to earthquake preparedness and response."
The world of huge parallel computers on the petascale has arrived. If past is indeed prelude and speed increases continue at current rates, within a decade, at least half of the world's 500 fastest computers will probably be petascale.
Access to such processing power will allow researchers in the health and life sciences to engineer proteins down to the atomic level, opening new doors to the treatment of several types of diseases. Climate analysis is another key field where petascale simulation will lead to much better modeling, enabling science-based approaches to emissions policy or to predicting the effects of global warming on air quality, agriculture, wildfires, and water supplies.
Scientists studying energy production and efficiency will also get new tools, permitting heretofore over-complex modeling of turbulence conditions and other factors that determine fuel efficiency, for instance, or the design of bio-fuels.
Before these new giants can be fully exploited, some big challenges
must first be addressed. UC Berkeley computer science professor
Katherine Yelick is working with colleagues in the Parallelism Lab to
bring such CITRIS-type applications and the petascale hardware and
systems software together.
UCB computer science professor Katherine Yelick."We are trying to expose the best features of the underlying hardware to the software," says Yelick. "The
hardware designers are trying to innovate and put in fast networks or
networks with very interesting connectivity patterns, and we want to
take full advantage of that," she says.
Yelick has one foot in the world of system-level software and the other in that of hardware development, which makes her particularly valuable to the coordination effort. She and her team have developed new compilers and programming languages (one based on C and another based on Java) for the new petascale computers.
One big challenge is the problem of pacing and managing the information flow through hundreds of thousands of processors. "It is like trying to get a million people coordinated and doing their jobs at exactly the same time," says Yelick.
Petascale machines not only have more chips, but each chip has more
processors than earlier generation supercomputers. Coordinating the
flow and sharing of so much activity is a job requiring new algorithms
and new approaches to applications programming, too, says Yelick.
This is a big problem because the work the computer is trying to do is not equally distributed among all of its processors. In modeling weather, for example, the Earth's surface can be divided into equal sized parts, and each given a dedicated processor. But if there is a hail storm somewhere, for example, there will suddenly be a lot of significant activity in the processors associated with those parts of the model. If the rest of the system has to wait for the processors working on the hailstorm, it can lose a lot of time, says Yelick.
In addition to such load imbalance issues, the team is working to minimize the time it takes for information to travel around these computers, some of which can be as big as a tennis court.
"Light travels pretty slowly," explains Demmel. "if processors on opposite sides of the computer have to send huge amounts of information back and forth, the time adds up fast."
Racks of servers at UCSC.
While Yelick straddles the gap between the systems-level programming and hardware design, Demmel straddles that between the applied math and the applications-level programming. "People who can work across one or more of those boundaries are very important in making these kinds of projects hang together," says Yelick.
People like Yelick and Demmel are finding themselves thrust from the rarified theoretical atmosphere of the high-end research computer world to what will soon be the center of a revolution in personal computing. As personal computers are forced to embrace parallel processors, they will face some of the same challenges as these high-end scientific computers.
In addition to the NSF bid to design and host a new peta computer for LLNL, Demmel and Yelick are just now completing another proposal for an Intel- and Microsoft-funded center for studying parallel computing applications for personal computers, games, hand-held devices and other commercial products.
"Now that Moore's Law can no longer be met by making single chips faster, everything is going to have to be parallel," says Demmel. "The computer industry will hit a wall unless it figures out how to deal with large-scale parallelism."