Re: [rdiff-backup-users] Compressed Backups? [repost]

[Maarten Bezemer]

Hi,

On Tue, 22 May 2007, Dan Mensom wrote:

> I would expect my usage case is a common one: I want to back up a lot of
> web + email content that should be very compressible. But from the first
> time I run rdiff, all that data seems to be stored uncompressed, taking
> 4-5x (or more) what it would take if I just created a tar.gz archive..

Now that's the catch!
The first time you run rdiff, all data will be stored uncompressed. So,
indeed maybe taking a few times more than it would take with a .tgz.
But, with each increment, only the changes will be stored, and they will
be stored in a .gz file. There's the big advantage: if you keep history of
over a month (of daily backups), and the mailbox is steadily growing, you
end up with storing the actual size, plus some small increment files. If
you'd like to keep that history with .tgz files, you end up with 30 times
the zipped archive, being about 30 / 5 = 6 times the amount of space
actually used on the source file system.

So, if you only want to keep history for one day, tgz is the way to go. If
you've got users that only report lost files a few days after they loose
them, rdiff-backup with small increments will probably be the better way
to go.

To answer your original question: to generate an increment-file,
rdiff-backup needs to have an uncompressed version of the file to compare
to. Although it could be done to store the backup in .gz, that would mean
unzipping and re-zipping all files during all backup runs. That would be
too CPU-intensive.


HTH,
 Maarten