View materialization for nested GPSJ queries

View materialization is a central issue in logical design of data warehouses since it is one of the most powerful techniques to improve the response to the workload. Most approaches in the literature only focus on the aggregation patterns required by the queries in the workload; in this paper we propose an original approach to materialization in which the workload is characterized by the presence of complex queries which cannot be effectively described only by their aggregation pattern. In particular, we consider queries represented by nested GPSJ (Generalized Projection / Selection / Join) expressions, in which sequences of aggregate operators may be applied to measures and selection predicates may be formulated, at different granularities, on both dimensions and measures. Other specific issues taken into account are related to the need for materializing derived measures as well as support measures to make algebraic operators distributive. Based on this query model, an efficient algorithm to determine a restricted set of candidate views for materialization, to be fed into an optimization algorithm, is proposed. Finally, the effectiveness of our approach is discussed with reference to a sample workload.