- Roadrunner is the fastest supercomputer in the world. IBM developed Roadrunner for the Los Alamos National Laboratory in New Mexico. The system cost $100 million to develop. When Roadrunner debuted in June 2008, it was the first computer to break what has been called the petaflop barrier -- the ability to operate at a rate of a thousand trillion floating point operations per second (FLOPS).
In testing, Roadrunner demonstrated 1.026 quadrillion calculations per second, as recorded by the Linpack measurement system. To illustrate its speed in everyday terms, IBM said that it would take 100,000 high-performance laptop computers to equal Roadrunner's processing power. In another comparison, the company said it would take the entire six billion people on the planet, working on hand-held computers (24 hours a day, seven days a week) 46 years to equal what Roadrunner can do in a single day.
[Image credit: Gizmowatch ]
Roadrunner's primary application will be work involving nuclear weapons, such as simulation of a nuclear attack. However, its level of processing power makes the supercomputer ideal for simulating complex interactions in many industries. Other potential applications include measurement of nuclear decay, studying complex weather patterns and developing capital market projections or astronomical calculations. Nanotechnologists and biochemists could explore intricate protein folding simulations or molecular interactions to design new drugs or cellular machines. Movie makers and information scientists could create previously unexplored levels of 3D rendering, both for entertainment and potential real-time modeling of cancer, surgical or other biomedical applications. Oil and gas producers could simulate underground geographies, determining the location of unexplored reserves.
Roadrunner, like other modern supercomputers, uses cluster computing to achieve higher and higher levels of performance. This approach is both cheaper and more extensible than the supercomputers of the early 90s, using commodity parts to achieve the final machine. Blue Gene, for instance, has been significantly upgraded since its first assembly. Consumers and research institutions with more limited means have used a similar approach to construct Beowulf clusters.
Roadrunner uses a hybrid design that combines more than 16,000 AMD Opteron cores, connected by InfiniBand with an equal number of Cell processors. The processors are housed in IBM Cell blade servers. Roadrunner runs the Red Hat Linux operating system. The supercomputer has a footprint of approximately 12,000 square feet. Roadrunner's hybrid architecture is expected to be duplicated and widely used by other cluster computing designers in the future.
The architecture of the processors themselves, developed by Sony, Toshiba and IBM, is identical to that used in the PlayStation 3 game console, the STI Cell Broadband Engine (Cell B.E.). The generation of chips used in Roadrunner are capable of two double precision calculations in 128 bit registers at full speed, with a peak of 100 gigaflops. Standard processing applications are routed to the Opteron processors, while more mathematically or CPU-intensive elements are routed to the Cell BE processors.
Roadrunner was named for the state bird of New Mexico. The speedy bird of the same name is colorfully depicted in a classic Warner Brothers cartoon. You can hear that old theme song here:
 |
Learn more about Cloud Computing, grid and utility computing |
| Learning Guide: Grid computing: This learning guide puts all grid computing related news, advice, whitepapers, and other resources at your fingertips. |
| High performance computing fast guide: In this guide to implementing high performance computing (HPC), find news, tips and advice on data center applications for supercomputing. |
| PNNL seeks supercomputer to battle radioactive plume: Pacific Northwest National Labs is replacing its four-year-old supercomputer. Its replacement will be used to tackle environmental remediation and sustainable energy. |
| Does the mainframe play a role in grid computing?: A grid computing effort that includes the mainframe can focus on improving grid robustness and introducing mission-critical applications to the grid. |
| High performance computing fast guide: In this guide to implementing high performance computing (HPC), find news, tips and advice on data center applications for supercomputing. |
| A grid glossary: Top 30 terms: This page puts the top 30 grid-related terms into one easy glossary. |
| Learning Guide: Grid computing: This learning guide puts all grid computing related news, advice, whitepapers, and other resources at your fingertips. |
| Grid computing: Our latest featured topic directs you to a learning guide we've created that puts the latest grid computing related news, advice, whitepapers, and other content at your fingertips. |
| Quiz: What's your grid grade?: Are you a grid groupie or just in a grid lock? Take this quiz to assess your knowledge of grid computing and all its related verbiage. This quiz is a tough one with lots of terms that have similar ... |
| CONTRIBUTORS: |
Mad New Mexico (Gilbert Mireles) |
| LAST UPDATED: |
11 Dec 2008
|
 |
Do you have something to add to this definition? Let us know.
Send your comments to techterms@whatis.com
|

 |
More resources from around the web:
|


');
// -->


 |
 |
|  |
RELATED GLOSSARY TERMS
| Terms from Whatis.com − the technology online dictionary |
 |
Blue Gene
(SearchDataCenter.com)
Blue Gene is an experimental parallel processing supercomputer developed by IBM that employs thousands of processors, each of which demands minimal...
|
 |
cluster computing
(SearchDataCenter.com)
In computers, clustering is the use of multiple computers, typically PCs or UNIX workstations, multiple storage devices, and redundant...
|
|

|