http://liw.fi/obnam/ondisk/ http://liw.fi/obnam/ondisk/ attempts to answer that in some detail. In summary: Obnam splits file data into chunks and stores each c

http://liw.fi/obnam/ondisk/ http://liw.fi/obnam/ondisk/ attempts to answer that in some detail. In summary: Obnam splits file data into chunks and stores each chunk only once. It looks up chunks using checksums, and does not need to download the chunk data again (unless you turn on verification mode to protect against checksum collisions). Obnam does download B-tree nodes, though, but they should be a fraction of the size of the file content data.