Eventzeit

Eventzeit is a term used in data processing and event-driven systems to denote the time at which an observed event actually occurred, as recorded within the event data. It is used to distinguish the real-world timing of events from other notions of time, such as when an event is processed by a system or becomes visible to storage.

In practice, event time is the timestamp carried by the event itself or generated by the event’s source. This semantics is essential for accurate temporal analysis, especially for time-based windowing, aggregations, and historical analytics. It is particularly important when events can arrive out of order or experience delays, because analyses based on event time aim to reflect real-world timing rather than the sequence of system processing.

Streaming platforms and frameworks often emphasize event-time semantics. Systems like Apache Flink and Apache Beam rely on event time to produce correct results under late-arriving data, using mechanisms such as watermarks to signal progress in event time and to handle late events.

Challenges associated with event time include clock skew across distributed sources, unsynchronized clocks, and late data that arrives after predefined windows have closed. Addressing these issues typically involves reliable timestamping at the source, clock synchronization practices, and carefully chosen windowing strategies with allowances for lateness and data out of order.

Etymology: Eventzeit is the German expression for “event time,” used in technical contexts to describe the real occurrence time of events. It is commonly contrasted with processing time (the time of observation by the system) and ingestion time (the time data enters the processing pipeline).

See also: event timestamp, processing time, watermarks, windowing, stream processing.