Database Performance for GIS : A Comparison of Database Schemas for Measurements with Spatial Attributes

Abstract: Spatial databases are becoming more and more important. One type of spatial database is a database of measurements, which combine positions with alphanumerical attributes. While these databases are important, little research has been done on them. There is therefore need for research on how to store this type of data most effectively.In this thesis a read performance benchmark and an execution plan analysis of three different database schemas using two different datasets of different size are presented. The three schemas investigated are the existing schema of the dataset, a schema where the spatial data have been split into a separate table, and a star schema. Additionally, these benchmarks were performed in two database management systems, Oracle Spatial and PostGIS.The results show that the choice of database schema has a real impact on performance, but none of the designs performs best in all cases. Star schema seems promising and may perform better than the existing schema in some cases, but more research is required to investigate further the specific cases where one schema outperforms the other. Finally, the results also show that PostGIS might perform better than Oracle Spatial for this type of data.