Kafka Streams threads vs. instances

[LL]

I am trying to understand/figure out best practices for scaling a Kafka Streams app. Reading the documentation here, it sounds as though threads and instances are fairly interchangeable as far as task distribution goes. What are situations where one would configure the number of threads in one instance rather than spin up another instance, and vice versa? Is it better to start up multiple instances or have a single instance running multiple threads?

[Eno Thereska]

There is a blog that goes over this in some more depth: https://www.confluent.io/blog/elastic-scaling-in-kafka-streams/

Eno