Event Time Windows Example
This example shows how to use event time windows in Apache Flink
The simple example creates a synthetic stream of sensor readings from a set of sensors. The readings are created roughly every 5 seconds, but have a have varying delay, which causes them to arrive out of order and delayed.
The stream is analyzed in three ways simultaneously, printing the results to stdout:
Low-latency event-at-a-time filter: Readings above a certain cause immediately alerts (latency of milliseconds afterevent arrival).
Processing Time Windows: Getting the maximum of all sensors in a group of sensors every 5 seconds (based on event arrival).
Event Time Windows: Computing average reading per sensor per minute (sliding) based on event timestamp. The windows have a delay of a few seconds, because they are only computed once the event time watermark signals that all relevant events have been received.