A Pipeline Processing Approach To GIS
Michael Koutroumpas, Chris Higgins · 2008
Much effort has been expended in the past researching parallel processing in GIS. However, uptake has been slow as the information architectures used were not based on interoperability standards and generally addressed only a small subset of GIS operations. Most research has concentrated on parallelizing specific algorithms for large-scale deployment rather than providing generic solutions. In this paper, based on open interoperability standards from both the Grid and Geospatial communities, we propose a pipeline processing approach to access, process and deliver geospatial data. The architecture is designed to be generic enough to be applicable in a wide range of spatial data infrastructure scenarios. We avoid the control and data dependencies that can hinder parallelization by taking into account the characteristics inherent in geographical data processing. GIS operations can then be modeled as a collection of processes arranged so that the output of each one is the input to the next. Using a feature based approach, a parallel processing pipeline can be constructed that overcomes problems imposed by dependencies between its processes. This paper describes an architecture based on widely available Grid middleware uniquely integrated with Open Geospatial Consortium (OGC) interface specifications. We further examine the requirements that should apply in the data stream between the processes to enable their parallelization. The viability of the model is demonstrated by applying it to selected GIS operations.