interesting
can you try to do a perfomance test while using block storage ?
[image: 1762419476977-81709355-5c09-4581-a424-9245a367bbdb-image.png]
This will store backup are multiple small ( typically 1MB ) files , that are easy to deduplicated, and the merge process will be moving / deleting files instead of modifying one big monolithic file per disk. It could sidestep the hydratation process.
This is the mode by default on S3 / azure, and will probably be the mode by default everywhere in the future, given its advantages
(Note for later : don't use XO encryption at rest if you need dedup, since even the same block encrypted twice will give different results)
