Performance Guideline

Hello community,

After some investigation, we identified a bottleneck in our camunda environment (7.21): we have a multi-instance subprocess with hundreds to thousands of iterations, each spawning an independent process instance to register credit transactions. This creates massive database contention and overhead.
Inspired by the recommendation in this thread to offload large batch iterations: I am considering removing the loop from Camunda entirely, pushing the batch payload into Redis, and having an external worker process the transactions sequentially/in controlled chunks before signaling the parent process when complete.

In other hand we are thinking to migrate to operaton, operaton has some guideline in this kind of use cases?

(Reference: Huge number of executions - Discussion & Questions - Camunda Forum)