fixedwindow

Fixed window is a windowing strategy used in stream and batch data processing in which the input data stream is partitioned into non-overlapping, fixed-duration intervals. Each interval, or window, has a defined start and end time, and every event is assigned to exactly one window based on its timestamp. For example, with a 5-minute fixed window, events with timestamps from 12:00:00 to 12:04:59.999 belong to the 12:00–12:05 window.

Within each window, computations such as counts, sums, averages, or other aggregations are performed independently. The results for a window are typically emitted when the window closes or when a configured trigger fires. Fixed windows are non-overlapping by design, which contrasts with sliding windows that overlap and move forward in smaller steps, and with session windows that have variable length based on activity.

Common use cases for fixed windows include generating regular interval metrics, such as per-minute or per-hour counts of events, aggregating sensor readings over uniform time blocks, and producing time-based dashboards. The simplicity of fixed windows makes them predictable and efficient for systems with uniform data flow.

Implementation considerations include how to handle late data and out-of-order events. Watermarks and allowed lateness are often used to determine when a window is considered complete and whether late elements should be included or discarded. Window alignment can be anchored to epoch boundaries or offset to match business calendars.

Related concepts include tumbling windows (often used interchangeably with fixed windows in many systems), sliding windows, and various trigger and watermark mechanisms that govern when results are emitted.