Maintaining data cubes under slowly changing dimensions

In the multi-dimensional model of data warehouses, data drawn from multiple sources is organized into fact tables and dimension tables. Data cubes pre-compute the aggregations of measurements along combinations of dimensions to speed up the processing of Online Analytical Processing (OLAP) queries. When the underlying fact tables or dimension tables change over time, data cubes need to be incrementally maintained to reflect these changes. In this thesis, we study the problem of incrementally maintaining data cubes under slowly changing dimensions. We argue that temporal features are essential in a data warehouse to avoid loss of historical information, and to help maintain data cubes to be consistent with the state of the underlying tables. We first characterize view maintenance problems based on changing dimensions. We then introduce our temporal multi-dimensional model that can describe changing dimension data, and we discuss strategies for incrementally maintaining data cubes based on that temporal model. A maintenance algorithm is developed, and its performance is evaluated experimentally. The results show that our approach is highly effective and is more efficient than re-materialization in some change cases. The results also show that our approach significantly outperforms existing incremental maintenance approaches