Azure Data Factory building blocks: pipelines, activities, datasets, linked services

• Linked service: the connection to a system, much like a connection string. One for your ADLS Gen2 storage account, one for Azure SQL Database, one for Databricks. Secrets such as passwords are usually read from Azure Key Vault instead of being typed in.
• Dataset: a named pointer to data inside a linked service, such as a folder of CSV files or a table. It describes where the data is and what shape it has.
• Activity: one step of work. Copy moves data, Lookup reads a small result, Stored Procedure runs SQL, Databricks Notebook runs a notebook, Execute Pipeline calls another pipeline.
• Pipeline: a group of activities that run together as one unit, with dependencies between them.
• Trigger: what starts a pipeline run, for example a schedule or a file arriving.

Trigger --starts--> Pipeline
                      |
                      +-- Copy activity
                            source: Dataset (orders.csv)  --uses--> Linked service (ADLS Gen2)
                            sink:   Dataset (dbo.orders)  --uses--> Linked service (Azure SQL)

A concrete example: load yesterday's orders file into a SQL table. You create two linked services, two datasets, one Copy activity inside one pipeline, and one schedule trigger. The work itself runs on an Integration Runtime, the compute that ADF uses behind the scenes.

Datasets and pipelines can take parameters, so one dataset can point at a different file each run instead of you creating hundreds of near-identical datasets. Expressions such as @pipeline().parameters.runDate fill the values in at run time.

Copy only moves data. For heavier transformation, ADF either uses mapping data flows or hands the work to Databricks or Synapse.

Interview tip: Say the chain out loud: "A trigger starts a pipeline, the pipeline runs activities, each activity reads and writes datasets, and each dataset lives in a linked service." Microsoft now steers new projects toward Data Factory in Microsoft Fabric, but ADF is still common in existing Azure estates, so interviews still ask about it.