On Tuesday, 08.04.2003 at 14:55 +0200, Toralf Lund wrote:

> A question a bit on the side:
> 
> I've been looking for a new drive or autoloader, and I notice that
> some vendors now assume a hardware compression ratio of 2.6:1 in their
> advertising material. Which leads me to the question in the subject;
> has anyone ever compressed real data that much? Or even achieved a
> ratio of 2:1, which is also frequently assumed? (In conjunction with
> backups, that is.) The best I've done with Amanda is about 1.8:1, but
> that's for software compression, of course.

The correct answer to the above is:  It Depends(tm)

There is no generally accepted 'standard' for what constitutes 'real
data', as you put it.

In particular, it depends on the nature of the data that you're backing
up.  Looking at our AMANDA reports, compression between 5:1 and 10:1 is
quite common for disk volumes where we have lots of 'data' - ASCII text
files that are quite repetitive, and therefore compress easily.  If the
disk volume consisted of just *.ogg and *.mp3, then your compression
ratio could be _less_ than 1:1, because no further compression is
possible and because of the overhead of actually storing the files.

Basically a 'vendor assumption' of _any_ compression ratio (low or high)
is worthless.  It is, after all, the native capacity that's important.

Which of the following makes more sense:

1. This tape has space for 40GB of stuff on it.

2. This tape has space for 80GB of stuff on it, if your files just
happens to be exactly compressible by a factor of 2 to 1 ...

Dave.
-- 
Dave Ewart
Dave.Ewart AT cancer.org DOT uk
Computing Manager, Epidemiology Unit, Oxford
Cancer Research UK
PGP: CC70 1883 BD92 E665 B840 118B 6E94 2CFD 694D E370