March 7, 2007

5:30 PM - 6:30 PM

Level: Intermediate

The debate about whether to use a star or a snowflake approach to dimensional modeling dates back to the early days of data warehousing. Increasingly sophisticated database systems and tools and faster hardware are changing the tradeoffs between performance and flexibility.

I will present the results of real-world tests comparing the performance of identical data marts implemented with both star and snowflake schemas.

This will feed into a framework for choosing which modeling approach is best for a given requirement. In particular, we will examine how the star/snowflake decision is driven by:

• Load and query performance
• Data model usability and maintainability
• Aggregate design and flexibility
• Ability to conform dimensions
• Tool selection
  

Tom Victory is a Principal Consultant at Decisionpath Consulting. His practice focuses on the delivery of BI and data warehouse solutions driven by business needs