Do Rule-Based Approaches Still Make Sense in Logical Data Warehouse Design?

As any product design, data warehouse applications follow a well-known life-cycle. Historically, it included only the physical phase, and had been gradually extended to include the conceptual and the logical phases. The management of phases either internally or intranally is dominated by rule-based approaches. More recently, a cost-based approach has been proposed to substitute rule-based approaches in the physical design phase in order to optimize queries. Unlike the traditional rule-based approach, it explores a huge search space of solutions (e.g., query execution plans), and then based on a cost-model, it selects the most suitable one(s). On the other hand, the logical design phase is still managed by rule-based approaches applied on the conceptual schema. In this paper, we propose to propagate the cost-based vision on the logical phase. As a consequence, the selection of a logical design of a given data warehouse schema becomes an optimization problem with a huge space search generated thanks to correlations (e.g. hierarchies) between data warehouse concepts. By the means of a cost model estimating the overall query processing cost, the best logical schema is selected. Finally, a case study using the Star Schema Benchmark is presented to show the effectiveness of our proposal.