The main objectives covered in this workpackage are:
- Adaptive data organization: Geospatial operations exhibit a high degree of locality in access which may be translated into physical data layout by clustering identifiers of geometries and their thematic content by geospatial location. This means that data access patterns will be local, resulting in improved CPU-memory traffic patterns and in decreased inter-node communication on clusters.
- Integration of complex geospatial algorithms into the database kernel: Examples of these would be route planning or map comparison/conflation. These typically involve graph-like traversal of linked structures, e.g. road network with highly local and highly repetitive access patterns, thus benefiting from a special memory-based data representation for often reused intermediate results.
- GeoSPARQL standard compliance: The existing implementation is very close to GeoSPARQL, thus this is mostly a matter of renaming functions and predicates.
- Data modelling with possible engine-level support for provenance or versioning: We expect such meta-information to be necessary in a heterogeneous melting pot of geodata.
Along with these objectives we will ensure fast searching, retrieval and display of the spatial data over the web. This involves different zoom levels, both in a spatial sense and in a semantic sense, i.e. hierarchical searches based on e.g. tourism taxonomies, depending on the use cases. Accurate query cost estimation is a key for this functionality, along with possible reuse of intermediate results.