Cost-Benefit Analysis of Cloud Computing versus Desktop Grids.PDF

Cost-Benefit Analysis of Cloud Computing versus Desktop Grids 1 1 1 1 2 Derrick Kondo , Bahman Javadi , Paul Malecot , Franck Cappello , David P. Anderson 1INRIA, France, 2UC Berkeley, USA Contact author: derrick.kondo@inria.fr Abstract Volunteer Computing (VC) platforms are another cost- efficient and powerful platform that use volunteered re- Cloud Computing has taken commercial computing by sources over the Internet. For over a decade, VC platforms storm. However, adoption of cloud computing platforms have been one of the largest and most powerful distributed and services by the scientific community is in its infancy computing systems on the planet, offering a high return on as the performance and monetary cost-benefits for scien- investment for applications from a wide range of scientific tific applications are not perfectly clear. This is especially domains (including computational biology, climate predic- true for desktop grids (aka volunteer computing) applica- tion, and high-energy physics). Since 2000, over 100 sci- tions. We compare and contrast the performance and mon- entific publications (in the world’s most prestigious scien- etary cost-benefits of clouds for desktop grid applications, tific journals such as Science and Nature) [15, 5] have doc- ranging in computational size and storage. We address the umented real scientific results achieved on this platform. following questions: (i) What are the performance trade-