Reporting on DBT-3 Database Benchmark
An example scenario is the one modeled by the DBT-3 database benchmark, available in open source at http://osdldbt.sourceforge.net/.
This benchmark defines a data warehouse with customers placing orders consisting of multiple line items. Each line item is a number of parts; parts are supplied by multiple suppliers. The benchmark comes with a data generator (dbgen) and set of queries (22 reporting queries and 2 update streams). The size of the database can be scaled using a scaling factor (SF). The default SF is 1, which represents 1 GB.
The schema contains the following table and foreign key relationships:
The largest table is "lineitem" (indicated in the drawing by the most saturated color). The number of tuples in each table is shown in parentheses. Arrows indicate the one-to-many relationships between tables.
This benchmark tests a number of business intelligence (BI) features, including:
• Joins between a large fact table (lineitem) and smaller dimension tables (part, supplier, nation, region), followed by groupings
• Large joins (between lineitem, order, and partsupp)
• Correlated subqueries
• Complex string predicates (such as LIKE)
• Top-N processing (getting the first N tuples of an ORDER BY)