Magnus Box backs up data by first splitting it into variable-sized chunks, which are individually compressed, encrypted, and uploaded. Magnus Box uses data-dependent chunking, efficiently splitting a file into consistent chunks even in the face of random inserts.
A backup job consists of a list of files and which chunks would be needed to reconstruct them. Any successive incremental backup jobs simply realize that chunks already exist on the server and do not need to be re-uploaded.
This chunking technique has the following properties:
- Both the oldest and the most recent backup jobs can be restored with the same speed
- Duplicate data does not require any additional storage, since the chunks are the same ("deduplication")
- There is never any need to re-upload the full file, regardless of the number of backup jobs
- There is no need for the server to be trusted to decrypt data.