Differential file systems?

Rainer Zocholl UseNet-Posting-72636- at zocki.toppoint.de
Sat Jan 18 16:36:00 EST 2003


Is there any progress in "differential file systems"?

Pardon if i missed that.

Usage Example:
I make a backup(*) copy of a database file, 1GB each file, today.
File size is "fixed" as the structure is predefined.
There are only a few place where values where changed.
Most of it is static.
Tomorrow i do the next backup and try to hard link to save disc space.
But "hard link" will fail because only one single byte was changed.
rsync have had all information and did sent only
that one byte, but then generates 999,999,999 Byte "overhead".
That's no good for performance and disc usage, or?

Where do i have to look to find such a file system, 
that does "hard links" on, let's says, 4KB clusters/inodes?

Of cause disc space is not important anymore to day.
But to copy 1GB from one side of the same disc to the other end took time.
And: Backing up a Database daily with one 1GB file, will fill a
30GB disc in approx. 4 weeks...with mostly identical informations.
The problem using rsync is the relatively long time required 
to copy the file locally. 
"Crosslink" / "reuse" of inodes would be much faster.

(*) In the sense of "versioning". 
    protecting against a "rm *,bak" or "vi" ;-)

More information about the rsync mailing list