Window Processing
Window processing addresses a well-defined stream processing problem described in depth by the "The Dataflow Model" whitepaper. A window operation turns data streaming records into a group of finite records, also known as bounded context, defined by the window size computed by a watermark operation. Fluvio performs a window processing operation by chaining multiple operators to assign timestamps, group them by key, and apply custom operations.
While there are several types of windows, and Fluvio will eventually implement all of them, this preview will focus on two: tumbling window and sliding window.
Tumbling windows are equal-sized, continuous and non-overlapping windows. Each record is present in exactly one window.
Sliding windows are equal-sized, continuous and overlapping windows. Each record may be present in one or more window.
Sliding windows is currently in development.