Stream Window Aggregator: Tumbling vs Sliding

Big data isn't only about storing huge volumes — a lot of it never stops moving. Real-time pipelines have to summarise an unbounded stream of events using only a bounded slice of it at a time, which is exactly what a window does. This simulator renders that slice as a glass box riding a conveyor of incoming events: in tumbling mode the box fills from empty to full and then flushes once, discarding everything and starting clean; in sliding mode the box always spans the last window size seconds and re-emits its aggregate every slide step, so overlapping windows share data. A fraction of events are held back before reaching the aggregator to simulate network jitter — if one arrives after its window's watermark has already closed, it is dropped rather than silently corrupting a stale aggregate. Tune window size, slide step and stream rate to see how each reshapes the trade-off between responsiveness, smoothness and data loss.