Keynote 1: Using Clouds for Science

We examine clouds in both their Virtual Machine based formulations and for new programming models/technologies such as MapReduce, Hadoop(Yahoo) and Dryad (Microsoft). We describe why clouds will either complement or replace Grids in future scientific environments and lead to more attractive computing environments for a broad range of users. We compare in detail Hadoop and Dryad with MPI noting the latter's ability to support iterative computations is important. We show how CGL-MapReduce supporting iterative MapReduce operations can perform well in both MPI and MapReduce application areas. We use applications from particle physics and bioinformatics as well as linear algebra and statistics kernels.