Re: [qubes-devel] ANN: Fast incremental backups project

[airele...@tutanota.com]

Dec 21, 2018, 1:39 PM by tas...@posteo.net:

On 12/20/2018 09:40 PM, Marek Marczykowski-Górecki wrote:

As for borg, I'm not sure a heavy emphasis on deduplication is appropriate for many PC applications. Its a resource drain that leads to complex archive formats on the back end. And my initial testing suggests the dedup efficacy is oversold: Sparsebak can sometimes produce smaller multi-generation archives even without dedup.

The borg default chunk size does not dedup Qubes volumes well. You get much better dedup with a smaller chunk size, for example `borg create --chunker-params=11,24,16,4095 ...`

Borg's dedup makes it simpler to think about. The repository is a pool of chunks and each backup is like a set of pointers to chunks. When a chunk is no longer referenced by any backup, it is deleted from the repo.