rDistributor - Remote Distribution of Complete Application Environments

Large scale grid computing is critical to researchers in Nuclear Physics, High Energy Physics, Climate Modeling, Genomic Research, and many other scientific pursuits. Grid organizations such as the Open Science Grid (OSG), TeraGrid, and the Enabling Grids for e-Science (EGEE) have assembled thousands of systems across widely distributed computing facilities. However, the inherent heterogeneity of those assembled systems ¿ different operating systems, system libraries, compilers, etc. ¿ pose a significant challenge to researchers who wish to run their applications on any available grid resource. In Phase I, a large scale nuclear physics application from Brookhaven National Laboratory was packaged in a virtual machine format. The virtual machine was then run on a virtual grid provided by Argonne National Laboratory. Although Phase I proved the capabilities of virtual grid computing, the overall process required several manual steps and was closely tied to the virtual grid environment provided by Argonne. In Phase II the project will automate the manual steps in the current process, allowing researchers to go directly from application packaging to job submission on the virtual grid. Applications packaged as virtual machines will be able to ¿contextualize¿ themselves based on information provided by the grid environment, allowing a single, generic application image to be run anywhere on the virtual grid. Finally, application environments will be enhanced to measure and track their resource consumption so that they can report resource usage back to the grid environment. Commercial Applications and Other Benefits as described by the awardee: This new model of computing should provide users with greater flexibility in their distributed computing infrastructure, while increasing the utilization of those same resources. Users would be able to rapidly promote applications from development to production by reducing the amount of testing required to validate applications across multiple production platforms. Also, users would be able to quickly respond to changes in application demand by deploying more application images to available virtual infrastructure.