Distributing a geographic visualization system on a network using a parallel data approach
We present a data parallel object-oriented distributed model for visualizing large spatial datasets built upon cell based grid structures. The model is based on object inheritance and aggregation hierarchies and is applicable to a wide class of spatial problems. The dependence of the model's performance on load distribution and communication latency is studied in detail. We use geometric parallelization to divide the problem domain into sub-regions (similar to finite difference domain decomposition). Our hierarchical implementation facilitates spatial objectification improves performance for domain wide searches. Cartographic projections are used as the "domain decomposition" functions for direct application to geophysical applications.