On Fri, Jun 02, 2006 at 08:14:28AM -0700, Joe Donner (sent by Nabble.com) wrote:
> 
> Firstly - thank you for the replies.
> 
> Secondly - I'm now more confused than before!
> 
> Is my understanding correct then that I can conceivably use this tapetype
> definition, but that I probably shouldn't expect to be backing up more than
> approximately 160GB to a tape?  Or, more to the point, approximately 135GB
> in this case?  And that I should really disable compression (but that this
> isn't necessarily required) and accept the tape's capacity of around 160GB?
> 
> I'm new to Linux (Redhat ES 3 in our case) and Amanda, so wouldn't mind
> things being spelled out to me - if you wouldn't mind doing that.

Your tape holds approximately 160GB of data.  Not YOUR data, simply data.
Minus maybe marketing exageration and base 10 vs base 2 definitions of GB
(1000^3 or 1024^3).

Computer scientists have found a way to take YOUR data and pack it into
fewer bytes of data.  The drive has one of these built in, gzip is another
way to pack the data tighter and is typically used by amanda.  There are
many others.

How much of YOUR data can be packed into 160GB of data depends on several
factors.  Chief among them are the nature of the compressor algorithm and
the nature of your data.  Drive manufacturers guess at the nature of YOUR
data (how can they know?) and claim a 50% compression.  And, on my last
amdump run, the overall compression by gzip was 46%, close to their guess
at compressability.  However, my individual DLEs, across 6 systems, ranged
from 82% compression to just 3%.

The compressor algorithm used by most tape formats has one nasty feature.
If YOUR data are random or nearly so, then passing it through the compressor
actually increases the number of bytes, by as much as 20%.  You saw the
effect of this when you ran amtapetype which feeds random data to the drive.
The tape held about 160GB of hardware "compressed" data, but those 160GB
represented only 136GB of YOUR data.

HTH
-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)