DP-900 · 1 Describe core data concepts
Analytical workloads
Exam objective: Describe features of analytical workloads
Analytical processing, or OLAP, reads large volumes of historical data to support reporting. Data typically moves through a data lake, a warehouse and an aggregated model.
Analytical processing, often called OLAP, looks backward across large volumes of data instead of handling one transaction at a time. It is read only or read mostly, and it usually works from a snapshot of the data rather than its live state.
Data typically flows through stages. Operational data is extracted, transformed and loaded, or extracted and loaded first with transforms applied after, into a data lake. From there it lands in a data warehouse or lakehouse, where a relational schema supports queries. Some of that data is then aggregated into an OLAP model, also called a semantic model, which precomputes measures across dimensions so reports run quickly.
A common way to organize that flow in a lakehouse is the medallion architecture:
| Layer | Holds |
|---|---|
| Bronze | raw data, ingested as is |
| Silver | cleansed, deduplicated, standardized data |
| Gold | aggregated, reporting-ready data |
Platforms such as Microsoft Fabric and Azure Databricks bring these stages together in one place.
On the exam, look for OLAP, semantic model, data warehouse, data lake, lakehouse or bronze, silver, gold.
Key points
- Analytical systems are read only or read mostly, and they analyze a snapshot (or a series of snapshots) of historical data rather than live transactions.
- ETL transforms data before loading it. ELT, common in lakehouses, loads the data first and transforms it afterward.
- A data warehouse stores data in a relational schema optimized for queries. A data lakehouse adds that relational querying on top of a data lake's flexible file storage.
- An OLAP model, also called a semantic model, preaggregates measures across dimensions so reports run fast and can drill up or down, for example sales by date, customer or product.
- The medallion architecture organizes a lakehouse in three layers: bronze holds raw ingested data, silver holds cleansed and conformed data, and gold holds business-ready aggregated data.
Exam trap
OLAP does not mean outdated. Microsoft Learn now calls the same aggregated, preaggregated storage concept a semantic model, and Power BI semantic models are the example you will meet most often.
Check yourself
A pipeline extracts data from source systems and loads it into a data lake first, applying any transformations only after the data has landed. Which pattern does this describe?
Go deeper on Microsoft Learn
Checked against Microsoft Learn on October 1, 2026.