Parallel scripting using cloud resources
Abstract
We will develop and improve techniques for using the Swift parallel scripting language to leverage distributed cloud resources to execute scientific applications at a high degree of parallelism. We will test this on 4 applications: 1) protein structure prediction; 2) protein-RNA docking; 3) quantum-level simulation of glassy materials; 4) analysis of fMRI data
Intellectual Merit
This project will address the systems issues of dynamic resource aggregation, efficient data management without shared cluster filesystems, and scheduling across fluctuating application demand levels and resource availability levels.
Broader Impact
The project will advance the ease-of-use of distributed cloud resources and make them more available to more scientists with less programming effort.
Use of FutureGrid
To test applications in biochemistry and neuroscience using parallel scripting techniques to leverage and federate cloud resources.
Scale Of Use
We would like to test on 4+ locations aggregating about 1000-2000 compute cores.We would run tests sporadically over several months, phasing in about 4 applications as they become ready.