Implementing block copy of a large file?

Question

Backup

Good evening.
I make a backup of the file server using Veeam Endpoint Backup, as a result I get a large .vbk file (a little more than 2TB) and a number of small .vib files on a separate disk of the file server.
I would like to copy these files to the NAS and the second server.
At the moment, copying is implemented using robocopy.
The problem is that every time you create a full synthetic backup, you have to download a large .vbk file in its entirety.
Are there any means of block copying this file (copying only updated parts of the file)?
The first thing that comes to mind is DFS, but there are quite large overheads for VSS (in any case, they were in 2008R2, maybe something has changed?) and it cannot be friends with NAS.
Are there any solutions to this problem?
Thanks in advance.

Answer 1

Generally there is, but is there a guarantee that, for example, there was no "shift" in any file header, which nullifies the whole idea of ​​block copying. I wrote a similar application when, after downloading GTA 5 from a third-party server (it took more than a week with my Internet), it turned out that some files were damaged (we will probably turn off the lights during download), of course, downloading is not an option, I had to chemistry. The same application is also suitable for your case (at least for "testing"), it scans the folder with all its contents and divides each file into blocks of about 1 mb, for each block it calculates the checksum, which it saves to the "report" file, after that, based on this report, it calculates which blocks are missing in that folder and which blocks need to be replaced and generates a "patch" that is simply "
Also in theory (didn't check and don't really know how to check) bittorent sync can do the same thing.

Answer 2

I would recommend Bittorrent Sync if the NAS supported it.

Answer 3

rsync is designed for this. If the file name is the same and the data blocks inside end up in the same place, then rsync will calculate the checksums block by block and transfer only the changes.
Also, the difference between large binary files can be done using xdelta. Doesn't directly solve the problem, but suddenly it will come in handy.