A composite model in Power BI is a semantic model that includes entities defined in two or more storage engines:

  • Local tables loaded in memory (VertiPaq)
  • Remote tables stored in a relational database (DirectQuery over SQL)
  • Remote tables stored in a Tabular model (DirectQuery over AS)

The Composite models video course and white paper describe in depth how Power BI and Analysis Services manage the different types of composite models you can create. The goal is to understand the underlying architecture to make sure you make the right design choices to create models that can be maintained over time and provide good performance to your end users.

The content is organized into the following modules:

  • Introduction to composite models
    • Introducing live connections
    • Introducing composite models
    • Local entities and remote entities
    • Connecting to a subset of the entire model
    • Wholesale queries and retail queries
    • Creating calculated columns
    • Creating calculated tables
    • Loading tables in the VertiPaq model
  • Understanding the architecture of composite models
    • Local model, remote models
    • Chaining
    • Lineage tag
    • Measures
    • Calculated columns
    • Tables and relationships
    • Using calculation groups
    • ALLSELECTED in composite models
    • Security in composite models
  • Query architecture in composite models
    • Using relationships
    • Retail queries through measures and queries
    • Optimizing retail queries
  • Composite model examples
    • Budgeting and sales
    • Static segmentation and ABC analysis
    • Dynamic segmentation
    • Dynamic ABC classification
  • Conclusion