Cloud · Azure in Depth

Azure Data Factory pipeline triggers and Integration Runtimes

azureMediumazureadforchestrationtriggersintegration-runtime

Interview question

What is the difference between schedule, tumbling window, and event triggers in ADF, and when do you need a Self-Hosted Integration Runtime?

How to answer

The short answer

Azure Data Factory (ADF) uses triggers to initiate pipeline executions and Integration Runtimes (IR) as the compute infrastructure that executes data movement and transformation activities.

Trigger types and operational trade-offs

Choosing the right trigger depends on whether pipelines run on wall-clock schedules, process historical intervals, or react to arrival events:

  • Schedule trigger: Executes pipelines at specific clock times or calendar intervals, such as every night at midnight. It does not track historical state or automatically failed runs.
  • Tumbling window trigger: Slices time into contiguous, non-overlapping intervals. It features built-in self-dependencies where a window can wait for the prior window to succeed, handles automated retries, and allows parallel backfilling across historical date ranges.
  • Event-based trigger: Listens to storage events such as BlobCreated via Azure Event Grid, executing ingestion pipelines immediately when a producer lands a file in ADLS Gen2.
Output
Schedule Trigger:        Fires at 02:00 UTC (Wall-clock, no state)
Tumbling Window Trigger: Slices 01:00-02:00 (Stateful, backfill capable)
Event Trigger:           Fires on BlobCreated in ADLS Gen2 (Reactive)

Integration Runtime selection

The Integration Runtime represents the actual compute bridge used by ADF:

  • Azure IR: Fully managed, serverless infrastructure that connects to publicly accessible endpoints and Azure PaaS services.
  • Self-Hosted IR: Software agent installed on an on-premise Windows server or private Azure virtual machine. It is mandatory whenever ADF needs to copy data from databases behind corporate firewalls or private networks.