Skip to main content


Advanced features in RisingWave.

📄️ Watermark

Watermark is one of the core concepts in a stream processing system. Data streams are infinite, and the watermark is used to impose limitations on these infinite data streams, guiding the system to handle data more efficiently. Watermark messages are interleaved within the data stream, and each watermark carries a timestamp indicating that all data up to that timestamp has arrived. RisingWave allows users to define watermark strategies on the time columns of source and append-only tables. RisingWave injects watermarks into the stream based on the defined expressions. For subsequent data that arrives later than the watermark, RisingWave filters and discards it.

📄️ Temporal Filter

The data streams processed by RisingWave are infinite. However, in many cases, users are only interested in data from a recent period (such as the past 1 hour, 1 day, or 1 week). Storing and processing unnecessary expired data is often unnecessary and a waste of resources. In some other systems, users are often allowed to define a Time-To-Live (TTL) to eliminate expired data, but the definitions can be vague, and the specific elimination time is unclear. Users also find it challenging to precisely specify the target states to eliminate. RisingWave provides a temporal filter syntax in SQL to help users eliminate expired data.