Queensland researchers simplify grid computing
Queensland researchers simplify grid computing
Apr 20, 2005: Researchers from Queensland University of Technology (QUT) are making it easier to combine the power of idle computers to create cheap but powerful virtual supercomputers.
The technique, which is called cycle stealing, or grid computing, taps into the number crunching capacity of a computer when it's not in use. This method is used around the world to harness the equivalent power of a supercomputer much more cost-effectively when working on projects that require many millions of calculations to be done quickly, such as trying to find a cure for diseases like AIDS or cancer.
While such projects demonstrate the potential of cycle stealing, researchers at QUT's Centre for IT Innovation are devising generic cycle stealing software that should enable anyone to build a virtual supercomputer for projects that need more power than your average PC on an ongoing basis, rather than merely for one-off projects designed to solve particular problems.
"The premise is that anyone who needs more computing horsepower than their desktop computer provides, should be able to get it via cycle stealing," explains PhD researcher, Jiro Sumitomo.
"Idle desktops can be combined to form ad hoc cycle stealing networks that provide high performance without the high cost associated with 'real' supercomputers."
Supported by a grant from Microsoft, the researchers have developed a framework called G2, which supports the creation of secure cycle stealing networks on the Internet. It allows people to contribute computing power using nothing more than a web browser.
The researchers designed G2 to meet user expectations of performance, reliability, ease of use and safety, including the ability to support a wide range of applications.
"By making cycle stealing networks cheap and easy to deploy, and by supporting many types of parallel applications, we hope to make high performance computing accessible to everyone."
An online demo of G2 is available on the web at http://g2.fit.qut.edu.au. The researchers hope that a public release will be available in the next six months.
Related Article: