⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.

Streaming Analytics: Tumbling Windows, Watermarks & HyperLogLog

This simulation models the analytics pipeline behind real streaming services: an unbounded event stream is grouped into fixed-width tumbling windows keyed by each event's own timestamp, a watermark tracks how long the engine waits before finalizing a window, and out-of-order "late" events that miss that cutoff are dropped exactly as they would be in Flink, Beam, or Kafka Streams. A live HyperLogLog sketch estimates the number of distinct viewers in the currently open window from a tiny fixed-size register bank, and its estimate is shown next to the exact ground-truth count so you can see the approximation — and its error — in real time. Tune the arrival rate, window width, watermark grace period, and late-event probability to see the trade-off between latency and correctness that every real streaming system has to make.