维度建模的现代方法 - 在列式数据库中（重复）.pdf

1、A Modern Approach to Dimensional Modelling In a Columnar DatabaseTruls BergersenData+AI Summit by DatabricksSan Francisco,June 13th 2024About the presenter Truls Bergersen Lives in Oslo,Norway 23 years of experience in data warehousing and business intelligence Data modeling and data integration exp

2、ert Background from row based relational databases Working with Azure Databricks for 1.5 years Founder of start-up company Okeanos AS(Contracted as)Lead architect of the Norwegian Tax Administrations data warehouseDisclaimer This presentation is a compilation of my personal thoughts on the future of

3、 dimensional modelling.Examples are simplified.There is no silver bullet,so one method will not fit all purposes.40 minutes is only enough to scratch the surface of this topic.Agenda1.Recap of dimensional modelling2.OBTs3.The way forward using star schemasIntro Dimensional modelling originates from

4、a joint research project conducted by General Mills and Dartmouth University in the 1960s.1 Used in the 1970s by both AC Nielsen and IRI.1 In 1996 the book The Data Warehouse Toolkit by Ralph Kimball is published.1 The Data Warehouse Toolkit,3rdedition,p15.Star Schemas Fact tables Dimension tablesFa

5、ct TableDimensionDimensionDimensionDimensionDimensionStar schemas are implementations of dimensional models in a relational database.They consist of:Fact Tables Store the performance measures I.e.aggregable numbers such asQuantityAmount Reference to dimension tablesvia foreign keys Usually contains

6、many rowsand few columns Four types of fact tables:Transactional Accumulative snapshot Periodic snapshot HybridFACTMeasure#1Measure#2Measure#nForeign key to Dimension 1Foreign key to Dimension 2Foreign key to Dimension nDimension tables Descriptive data giving context to the facts.Has a primary key