Block Cloning is a method that allows for the cloning of a file or subset of data blocks without actually copying the data itself. Unlike automatic deduplication, Block Cloning requires the use of a dedicated system call to clone the desired file or blocks. This process operates exclusively on metadata, making it extremely fast. Some key differences between Block Cloning and deduplication include the fact that Block Cloning only creates an entry in its tables when there are at least two references to a given data block, whereas deduplication keeps all data blocks in its table. Additionally, Block Cloning does not require the passing of real data to calculate a hash, making it more efficient. Overall, Block Cloning offers a manual deduplication option that is faster and requires less space compared to traditional deduplication methods.
https://github.com/openzfs/zfs/pull/13392