Grid-based geo-middleware for geographic analysis: Theory, method, implementation, and evaluation

Computational Grids enable the coordinated sharing of large-scale distributed heterogeneous computing resources that can be used to solve computationally intensive problems in science, engineering, and commerce. Grid applications are made possible by high-speed networks and a new generation of Grid middleware that resides between networks and traditional applications. The purpose of this dissertation research is to investigate the theory, methods, implementation, and performance of Grid-based geo-middleware (hereafter designated geo-middleware) that enables the use of Grid computing resources for computationally intensive geographic analyses. In this dissertation, the spatial computational domain theory is established to bridge the gap between existing knowledge about spatial data parallelism and the computational intensity of geographic analyses. Through the application of this theory, geo-middleware methods are designed to include two major components: domain decomposition based on region quadtrees and space-filling curves; and task scheduling that uses static strategies to significantly improve performance and achieve efficient use of Grid resources. The methods are implemented and evaluated using two case studies (inverse distance weighted interpolation and the G*i (d) statistic) that were selected to validate the spatial computational domain theory. Based on a series of computational experiments, the performance evaluation demonstrates the feasibility of using the geo-middleware to achieve high performance solutions to computationally intensive geographic analyses. In addition to domain decomposition and task scheduling methods, several other geo-middleware components are prototyped as Grid services to create a problem solving environment---GISolve. GISolve is implemented to integrate three types of Grid services: application-specific, OGSA (Open Grid Service Architecture), and interaction-specific services using Grid portal technologies.