Overhead of CDC on Couchbase Server

My project uses Kafka Connect to capture database changes to a Couchbase bucket. It first gets an initial snapshot of the bucket and publishes all the docs to Kafka. After that, it captures all the new changes and publishes those changes as well.

Does anyone know about the runtime overhead the Connector adds to the CB server? I know this may depend on the size of the bucket and change rate for the initial snapshot and following streaming. Just wondering if someone has done some perf study or analyses so we can have a sense of it. Thanks in advance.

I’m not the expert here, but I assume it is very efficient as I believe the same mechanism is used for replication.