Economics of Cloud Computing: a Statistical Genetics Case Study

摘要

We describe an experiment which aims to reduce significantly the costs of running a particular large-scale grid-enabled application using commercial cloud computing resources. We incorporate three tactics into our experiment: improving the serial performance of each work unit, seeking the most cost-effective computation cycles, and making use of an optimized resource manager and scheduler. The application selected for this work is a genetics association analysis and is representative of a common class of embarrassingly parallel problems.