Handling each new item of data in real time as soon as it arrives, rather than gathering it into batches. It fits time-sensitive work, with latency measured in seconds or milliseconds against the hours batch processing usually needs.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Stream processing in context, with comparison tables and the common traps.
Terms in this definition
- Batch processing
An approach where incoming records are gathered and kept, then handled all at once, whether on a timetable, once a set amount has built up, or on another trigger. It has more delay than stream processing but copes well with large amounts of data.
Related terms
- IoT Central
An IoT application platform delivered as SaaS; it does not do stream processing. Microsoft is retiring it: new applications can't be created from 23 September 2026 and existing ones run until 20 September 2029, with IoT Hub, the Device Provisioning Service and Microsoft Fabric as the recommended path.
- Streaming data
Events that keep arriving without end, for example log entries, sensor readings or social media posts; stream processing is how they are handled.
- Streaming dataflows
Once offered no-code preparation and ingestion of streaming data in Power BI but has now been retired. Those capabilities live on in Azure Stream Analytics no-code stream processing.