Event Processing (one of the new capabilities of IBM Event Automation) makes it easy to perform stateful processing of streams of events from Kafka topics. In this post, I want to drill down a little into how windowed processing behaves.
I’ve enjoyed the chance to see the solutions that people have started to build with the tool. As part of this, I’ve been helping several people to understand the results produced by the event processing flows they’ve made.
These often started with different questions, such as:
- Why hasn’t my flow produced any results?
- Why isn’t my one-minute window producing one result every minute?
- Why did the last one event on my Kafka topic cause results to be produced for several different windows?
- etc.
However, these are often symptoms of a single common question: how windowed processing operations behave.
I’ve tried to come up with simple ways to demonstrate how it works, so in this post I want to share how I’m currently explaining it.