Automating Multiple Schema Generation using Dimensional Design Patterns
Monali A. Deshpande · OhioLink ETD Center (Ohio Library and Information Network) · 2009
A data warehouse is a repository where data is collected from various sources, integrated, and represented in a dimensional model.The structure and description of the data stored in the repository is called the schema.The two basic approaches for designing a data warehouse schema are data-driven and requirement-driven.Data-driven approaches use operational sources as the guide to creating schemas, while requirement-driven approaches are guided by end-user query and analysis needs.Most approaches in the literature are data-driven; however, some researchers have initiated research into requirementdriven methodologies.Jones and Song [JS05, JS07] propose a requirement-driven approach inspired by design patterns from software engineering.They define dimensional design patterns (DDPs) that capture features common to many dimensional schemas.The use of DDPs assists a designer in creating a dimensional schema.We automate the process of creating one or more star schemas using DDPs and perform case studies to illustrate use of the software tool in different domains.In addition, we automate a process to examine the generated schemas to identify shared dimensions, called conformed dimensions that can be further used by the designer to refine and merge the schemas.