Joins are NOT Expensive! - Raw Reading
Summary
Two data modeling approaches are tested for performance: a traditional dimensional model with joins vs a One Big Table, using DuckDB and PostgreSQL. The author shows that joins are not expensive and that wide tables can incur non linear CPU costs as the number of columns increases, with caveats based on engine and workload. Kimball's data warehousing principles are discussed, and the article concludes that joins often win, especially with filters.