An Aspect-Oriented Approach for Disaster Prevention Simulation Workflows on Supercomputers, Clusters, and Grids

Abstract

Abstract-Computer simulation is an important factor in today's disaster prevention procedures. Simulation codes assess the evolution and impact of various physical phenomena in domains such as nuclear and environmental sciences, and ultimately help saving lives. However, new and more computationally demanding models, and new regulations for personnel training have increased the demand for computational power. While the existing simulation codes can be ported to computing environments that can meet the new demand, such as supercomputers, clusters, and grids, it is too expensive and time-consuming to rewrite and re-certify them. Instead, in this work we propose an aspect-oriented approach that takes existing simulation functionality and combines it with functionality required to run the simulation on different computing environments transparently to the simulation developer. Through experiments in the DAS-3 multi-cluster grid we show that our approach increases the reusability, the maintainability, the scalability, and the robustness of a real disaster prevention simulation, while incurring a low performance overhead.

Reducing Time Delays in Computing Numerical Weather Models at Regional and Local Levels : A Grid-Based Approach

“…Other complex applications exist which also use distributed workflows, and require a large number of computing nodes, which are typically unavailable within one organization. The Disaster Simulation system [29] assesses the evolution and impact of various physical phenomena, in domains such nuclear and environmental science. Aircraft industries use large scale multi-physics applications (aircraft flight dynamics simulation) to reduce the time of development [30].…”