Star Schema vs. OBT for Data Warehouse Performance

Michael Kaminsky from Gradient Metrics conducted a benchmark comparison of data architectures, specifically focusing on the performance implications of different warehouse distribution patterns under normal BI-style workloads within Redshift, Snowflake, and BigQuery. The results showed that denormalized tables resulted in faster query response times compared to star schemas for all three warehouses. The speed improvement ranged from 25% to 50%, depending on the warehouse used. This analysis was conducted using a subset of TPC-DS benchmark data and aimed to understand how different data architecture patterns perform once a warehouse has been chosen, rather than comparing warehouses themselves.