Hey folks, I have a question about best practice. We have two types of IoT devices that record an event every 5 minutes but one type only sends them bundled every hour. They arrive from two streams, need to run though the same processing logic in Flink including some event time windowing, and get output into a single sink. The data is keyed by device ID and data from different devices does not interact. Is it possible to process the data in a single Flink job without delaying the faster events by having high allowed lateness? Would two deployments of the same job be the alternative?

Is it possible to process the data in a single Flink job without delaying the faster events by having high allowed lateness?

If you configure a high allowed lateness, all data will just be later (basically because Flink needs to be wait until it can be sure that no later data has arrived)

Thanks for the reply. So if we do not want to delay the faster events, we would need two jobs with different allowed lateness?

Yes