I'm trying to understand KSQL and whether it has the appropriate performance characteristics for our use case.
Use Case Given Thousands of IoT connected devices sending telemetry data, have a service that aggregates, ranks and sorts that data into various graphs, and returns that data to each device a real-time stream.
Does KSQL appropriate to this use case - especially when returning a different aggregate set per device?
Yes. No. It depends :-D
This is a pretty broad question that can only be answered broadly.
KSQL is built on top of Kafka Streams, part of Apache Kafka. As a distributed system you can partition data and scale out horizontally to parallelise processing. If you need to increase the rate at which you're processing streams of data you add in more KSQL servers and so long as your topic is partitioned appropriately the work will be distributed across all the additional servers.
