Application Plugins for Distributed Simulations on the Grid

Abstract

Computing grids are today still underexploited by scientific computing communities. The main reasons for this are, on the one hand, the complexity and variety of tools and services existent in the grid middleware ecosystem, and, on the other hand, the complexity of the development of applications capable to exploit the grids. We address in this work the challenge of developing grid applications that keep pace with the rapid evolution of grid middleware. For that, we propose an approach based on plugins for grid applications that encapsulate a set of commonly used type of grid operations. We further propose more complex high-level functionalities, such as the plugins for remote exploration of simulation scenarios and for monitoring of the behavior of end-user applications in grids. We provide an example of a grid application constructed with these software components and evaluate based on it the performance of our approach in the context of the simulation of biological neurons. The results obtained on test and production grids demonstrate the usefulness of the proposed plugins, with a small performance overhead compared to traditional grid tools.