This epic can be considered part of the work toward having netobserv in core payload, although it's not only useful in that case. Current users of netobserv can be interested in this mode even if they have installed the OLM operator.

The goal is to provide an intermediate deployment model between the Direct model (that deploys FLP as a DaemonSet) and the Kafka model (that is more scalable but requires Kafka).

The "Service" (or "Deployment+Service", not sure how to call it) is scalable (you can configure how many replicas you want), and doesn't require extra dependency like Kafka.

It should provide better performance than the "Direct" mode, especially on large cluster, by avoiding too much cache duplication between FLP instances.