Cloud Computing for parallel Scientific HPC ApplicationsFeasibility of running Coupled Atmosphere-Ocean Climate Models on Amazon’s EC2.PDF

Cloud Computing for parallel Scientific HPC Applications: Feasibility of running Coupled Atmosphere-Ocean Climate Models on Amazon’s EC2. [Extended Abstract] ∗ Constantinos Evangelinos and Chris N. Hill Department of Earth, Atmospheric and Planetary Sciences, MIT 77 Massachusetts Ave. Cambridge, MA 02139, USA ce107@ and cnh@ Abstract the model can also be a significant impediment to progress. In this article we describe the application of HPC standard In particular a queue based batch paradigm makes on- benchmark tests to Amazon’s EC2 cloud computing system, demand, responsive and highly customized computational science research hard to undertake. On-demand cloud com- in order to explore the utility of EC2 for modest HPC style applications. Based on these benchmarks we find that the puting appears to offer an attractive new dimension to HPC, EC2 cloud system is emerging as a credible solution for in which virtualized resources can be sequestered, in a form supporting responsive on-demand, small sized, HPC appli- customized to target a specific scenario, at the time and in cations. We illustrate this point with a demonstration of a the manner they are desired. We believe this concept lends low-order coupled atmosphere-ocean simulation running in itself to more interactive responsive computational science. parallel