Bridging the Gap between Desktop and the Cloud for eScience Applications

The widely discussed scientific data deluge creates a need to computationally scale out eScience applications beyond the local desktop and cope with variable loads over time. Cloud computing offers a scalable, economic, on-demand model well matched to these needs. Yet cloud computing creates gaps that must be crossed to move existing science applications to the cloud. In this article, we propose a Generic Worker framework to deploy and invoke science applications in the cloud with minimal user effort and predictable cost-effective performance. Our framework addresses three distinct challenges posed by the cloud: the complexity of application deployment, invocation of cloud applications from desktop clients, and efficient transparent data transfers across desktop and the cloud. We present an implementation of the Generic Worker for the Microsoft Azure Cloud and evaluate its use for a genomics application. Our evaluation shows that the user complexity to port and scale the application is substantially reduced while introducing a negligible performance overhead of < 5% for the genomics application when scaling to 20 VM instances.