The Drill Down Benchmark

Data Mining places specific requirements on DBMS query performance that cannot be evaluated satisfactorily using existing OLAP benchmarks. The DD Benchmark - defined here - provides a practical case and yardstick to explore how well a DBMS is able to support Data Mining applications. It was derived from real-life data mining tasks performed by our Data SurveyorTM tool running on a variety of DBMS backends. We describe initial results obtained using both the Monet system and a relational DBMS product as backend.