Developing Reliable Taxonomic Features for Data Warehouse Architectures

Since there is a large variety of data warehouse architectures with different structures and components, it is very difficult and time-consuming to systematically analyse them and obtain insights from those architectures. One effective way to understand those architectures is using a taxonomy to classify them. However, most of the taxonomic features are derived in an ad-hoc way and the reliability of those features is unknown. This paper therefore is to develop a set of reliable features by modeling different data warehouse architectures and further generate the structural knowledge represented by a taxonomy. This taxonomy is further validated by evaluating two real-world data warehouse architectures from IBM and Facebook.