Grid computing applied to the boundary element method

Over decades, since the invention of the first computers, hardware and software have been created or modified to follow the increasing complexity of scientific and engineering problems. However, since there is always a limit to the performance of a workstation, server or even a local parallel machine, programs can bypass these limitations by using remote resources with greater computational power. Grid computing is one of the most recent developments of computer science which allows engineers and scientists the remote use of geographically distributed heterogeneous resources such as high performance computers and supercomputers. This emergent paradigm makes possible to create a network of shared hardware and software through the Internet, owned and maintained by different organizations. Such computing infrastructure enables to solve a wide variety of problems that could not be addressed in a local environment. An well-known example of use of distributed computing power is SETI@home, a scientific experiment where worldwide home computers connected to the Internet are used to analyze radio telescope data in the Search for Extraterrestrial Intelligence. The same approach is used by the BBC Climate Change Experiment that gathers the processing power of thousands of home and office computers around the world to predict climate changes. The simulation of the flow of blood through human arteries, the effect of shock waves propagation in earthquakes and the gravitational effects of black hole collisions are also examples of grid enabled scientific applications. This paper introduces GridSolve, a software tool for grid computing developed at the Innovative Computing Laboratory of the University of Tennessee and presents its use in the implementation of engineering codes on this new environment. Since this technology is not widespread among engineering developers, the authors present this technique applied to a well-known Fortran program for the solution of two-dimensional elastostatic problems with the Boundary Element Method. The guidelines provided here may also be extended to other numerical methods.