modelgrid

This will be an experimental evaluation of software engineering methods in the "Grid world". At present it is a very difficult process to prepare an algorithm which runs efficiently on the grid. In contrast to established software engineering processes, there is no support of CASE tools for creating grid applications. Yet CASE tools are very promising to identify and model topics like CPU and memory constraints, throughput of network links, their reliability, capabilities of application components, their dependencies between each other and parallelization issues. Tool support, component based modeling, design patterns, and contracts could simplify the development process for grid applications significantly. For that reason there is need to explore how the development process of grid applications can be supported with tools. In this thesis a typical sample grid application - the DOUG (Domain Decomposition on Unstructured Grids) package - will be decomposed in components and visualized in modeling tools. The extracted components should be semantically tagged and be deployable to the grid. This work will lead to results how to decompose current grid applications and recompose them to better structured applications with reusable entities. Also strategies for tagging and application configuration will be evaluated from this approach. The successor of this work will be gridcomponents.