Skip to main content
- What it is
- An open-source engine for stateful stream processing — continuous computation over unbounded event streams, with batch as a special case.
- What makes it distinctive
- Event-time processing with watermarks, so out-of-order and late events are handled correctly.
- Managed application state with checkpointing for exactly-once semantics.
- Windowing (tumbling, sliding, session) for aggregations over time.
- Where it runs on AWS
- Amazon Managed Service for Apache Flink — serverless Flink applications over streams, formerly Kinesis Data Analytics. See Amazon Kinesis.
- Amazon EMR when you want to operate the cluster yourself.
- Exam signal
- “Real-time analytics on a stream”, “windowed aggregation”, “anomaly detection as data arrives” → Managed Service for Apache Flink.