Eco-data warehouse design through logical variability

Abstract

The database (DB) is one of the active communities dealing with the non-functional requirements (NFRs) when designing advanced applications. The fulfillment of the NFRs is usually performed along the phases of DB life cycle in an isolated way. The physical design phase took the lion’s share of these studies, because it is an important factor for a successful DB deployment in terms of performance metrics. By carefully analyzing these studies, we figure out that target DBs are assumed to be already deployed, meaning that their logical models are frozen. This assumption surely becomes questionable, since it ignores the chained aspect of the life cycle. Knowing that many variants of a logical schema may exist due to the presence of dependencies and hierarchies among attributes; it is worth studying the impact of this variation on the physical design. In this paper, we firstly identify the dimensions of the variability of a logical schema and their modeling. Secondly, we propose a methodology, by highlighting the efforts that designers have to make, to evaluate the impact of the logical schema variability on the physical design (by considering logical or physical optimization), where both energy consumption and query performance are considered. Finally, intensive experiments are conducted to evaluate our proposal and the obtained results show the real impact of variability on data warehouses (DW) eco-design.