Current San Francisco · 4 November 2026
Kafka Streams Needs Real Timers
Hartmut Armbruster
Records tell you what happened. Only a timer can tell you what didn't. Kafka Streams gives you punctuators; Flink gives you a timer per key. This talk compares the two, sketches a per-key timer API for the Kafka Streams Processor API, and shows the same idea running on StoatFlow.
Slides and handout
Get the StoatFlow 1.0 launch email
StoatFlow 1.0 is on its way. Leave your email and I'll tell you when it's generally available, and when you can start a trial without a call.
Timers in the StoatFlow docs
- Timers in the Processor API
- Named timers
- Event-time and processing-time timers
- Testing with TopologyTestDriver
- Configuring the timer backend
- Punctuators and timers, compared with Kafka Streams
- Timer, punctuator or scheduled source?
The slides stay as presented. The docs follow the current release.