During data warehouse design, the designer frequently encounters the problem of choosing among different alternatives for the same design construct. The behavior of the chosen design in the presence of evolution events is an important parameter for this choice. This paper proposes metrics to assess the quality of the warehouse design from the viewpoint of evolution. We employ a graph-based model to uniformly abstract relations and software modules, like queries, views, reports, and ETL activities. We annotate the warehouse graph with policies for the management of evolution events. The proposed metrics are based on graph-theoretic properties of the warehouse graph to assess the sensi tivity of the graph to a set of possible events. We evaluate our metrics with experiments over alternative configurations of the same warehouse schema. © 2008 Springer Berlin Heidelberg.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below