Apache Pinot

Hi guys, let me ask a question about <https://docs.pinot.apache.org/operators/operating-pinot/instance-assignment#replica-group-instance-assignment|Replica-Group Instance Assignment>, I'd like to better understand how it works (more on this thread).

Just to confirm: the idea is that each Replica Group (RG) has a set of servers, and each server in that RG has a certain number of segments available, but not all of them, and then the RG ensures that, together, the group of servers as a total can handle all segments. Is this correct?

So the idea is that each RG can handle any part of the data, but I can have many RGs so that I can have proper replication and partitioning of the data in multiple servers?

One replica group is one full copy of the data.

So a query will fanout to only one replica group.

Cool, and then the Broker will know which Servers, in that RG, will handle each part of the data, right?

It makes Pinot horizontally scalable for throughput. If you get 1k read qps from 1 replica group (say 3 servers), then you will get 10k read qps from 10 replica groups