FabricPool and Deduplication

As I understand, FabricPool tiering would not transfer deduplicated data over to AWS S3, and it would have to be rehydrated first. My question is:
1. Where rehydrating process would take place, on the storeage cluster?

2. then when retrieving data back to the sotrage cluster from S3, will data be duplicated again?
3. how much performance could be degraded?

Thanks!