We can create segments outside of Apache Pinot cluster. That's good. But I don't understand how it upload segments to offline server from deep storage. For example,
I write segment to Hdfs. Then I call segment tar push job. If I didn't understand wrong, segmenttarpushjob downloads created segment from deep storage(s3), then it uploads segments to the controller via rest api, and controller sends segments to offline servers.
Will not this process create bottleneck while sending segments to controller? What will happens if offline servers downloads segments from deep storage directly?