I was somehow under the impression when first looking at this software that the incremental backups were going to save storage space on the backup server. I found your project from links comparing it to rsync/rdiff-backup types of solutions.
However, it appears that while the generation of snapshots is block level (data sent from client to server), the actual space used on the backup server is a full backup for every snapshot. This is untenable for large databases or virtual machines. Is this correct? Are there plans for having it function more like rdiff-backup (i.e. reverse incremental backups)?
Yes, whole files are saved twice even if only a few blocks differ.
This can’t be changed without using a custom backup file format.
There is one exception: With btrfs on Linux one can use a trick to only save block differences between file backups (via reflinks). ReFS seems to be really lacking feature wise, so there’s not much hope in that area for Windows.
Correction, with Windows Server 2012 and data deduplication the same behavior should be possible.
Correction, with Windows Server 2012 and data deduplication the same behavior should be possible.[/quote]
Cool. Did not know that existed . I take it back. (Though it is more efficient with btrfs, because you get per backup snapshots and “deduplication” for free, plus real deduplication, but that is still in development.)