A Service-Oriented Architecture for Scientific Computing on Cloud Infrastructures

This paper describes a service-oriented architecture that eases the process of scientific application deployment and execution in IaaS Clouds, with a focus on High Throughput Computing applications. The system integrates i) a catalogue and repository of Virtual Machine Images, ii) an application deployment and configuration tool, iii) a metascheduler for job execution management and monitoring. The developed system significantly reduces the time required to port a scientific application to these computational environments. This is exemplified by a case study with a computationally intensive protein design application on both a private Cloud and a hybrid three-level infrastructure (Grid, private and public Cloud). Topics Parallel and Distributed Computing.