Storage and indexing of relational OLAP views with mixed categorical and continuous dimensions

Due to the widespread adoption of location-based services and other spatial applications, data warehouses that store spatial information are becoming increasingly prevalent. Consequently, it is becoming important to extend the standard OLAP paradigm with features that support spatial analysis and aggregation. While traditional OLAP systems are limited to data characterized by strictly categorical feature dimensions, Spatial OLAP systems must provide support for both categorical and spatial feature dimensions. Such spatial feature dimensions are typically represented by continuous data values. In this paper we propose a technique for representing and indexing relational OLAP views with mixed categorical and continuous data. Our method builds on top of an established mechanism for standard OLAP and exploits characteristic properties of space-filling curves. It allows us to effectively represent and index mixed categorical and continuous data, while dynamically adapting to changes in dimension cardinality during updates. We have implemented the proposed storage and indexing methods and evaluated their build, update, and query times using both synthetic and real datasets. Our experiments show that the proposed methods based on Hilbert curves of dynamic resolutions offers significant performance advantages especially for view updates.