DP-900 · 4 Describe an analytics workload
Data ingestion and processing
Exam objective: Describe considerations for data ingestion and processing
Ingestion moves data from sources into an analytical store, reshaping it along the way. ETL transforms before loading, ELT transforms after.
Before data can be analyzed at scale, it has to get from its original source, such as a database, a file, or a live event stream, into an analytical store such as a data lake or a warehouse. That move is called ingestion, and it almost always reshapes the data along the way: values get cleaned, filtered and restructured so the result fits a schema built for reporting rather than for day to day transactions.
Two load patterns differ only in when the transform happens.
| Pattern | Transform happens |
|---|---|
| ETL (extract, transform, load) | Before the data reaches the store |
| ELT (extract, load, transform) | After the data has been copied into the store |
Ingestion is not only a batch affair: a pipeline can pull in a nightly file drop and, separately, a continuous event stream, side by side. The heavy lifting is usually spread across a cluster of machines working in parallel, which is what lets a pipeline scale from a few rows to billions.
A pipeline itself is built from activities that read, move or transform data, connected to sources and destinations through linked services that hold the connection details.
On the exam, watch for "before loading" pointing to ETL and "after loading" pointing to ELT, and remember that ingestion covers both batch and real time sources.
Key points
- Ingestion moves data from transactional stores, files or event streams into an analytical store such as a lake or warehouse.
- ETL transforms the data before it is loaded into the store; ELT loads the data first and transforms it afterward.
- Ingestion covers both scheduled batch loads and continuous real time streams, often in the same solution.
- Processing is commonly spread across multi node clusters so it can scale to high volumes.
- A pipeline is built from activities, connected to sources and destinations through linked services.
Exam trap
ETL and ELT differ only in when transformation happens, not in what gets loaded. Assuming data lands in the store already transformed when a scenario describes an ELT process is the usual mistake.
Check yourself
For each statement about ingestion, select Yes if it is true. Otherwise, select No.
- In ETL, data is transformed before it is loaded into the analytical store.
- In ELT, data is transformed before it is copied into the store.
- Data ingestion can include both batch processing and real-time processing of streaming data.
Go deeper on Microsoft Learn
Checked against Microsoft Learn on October 1, 2026.