Abstract
Geoprocessing is commonly used in solving problems across disciplines which feature geospatial data and/or phenomena. Geoprocessing requires specialized algorithms and more recently, due to large volumes of geospatial databases and complex geoprocessing operations, it has become data- and/or compute-intensive. The conventional approach, which is predominately based on centralized computing solutions, is unable to handle geoprocessing efficiently. To that end, there is a need for developing distributed geoprocessing solutions by taking advantage of existing and emerging advanced techniques and high-performance computing and communications resources. As an emerging new computing paradigm, grid computing offers a novel approach for integrating distributed computing resources and supporting collaboration across networks, making it suitable for geoprocessing. Although there have been research efforts applying grid computing in the geospatial domain, there is currently a void in the literature for a general geoprocessing optimization.
In this research, a new optimization technique for geoprocessing in grid systems, Geoprocessing Optimization in Grids (GOG), is designed and developed. The objective of GOG is to reduce overall response time with a reasonable cost. To meet this objective, GOG contains a set of algorithms, including a resource selection algorithm and a parallelism processing algorithm, to speed up query execution. GOG is validated by comparing its optimization time and estimated costs of generated execution plans with two existing optimization techniques. A proof of concept based on an application in air quality control is developed to demonstrate the advantages of GOG.
|
