What you lose is time. Regardless of what v42 was written to do (yes, it
could in theory be applied to any data stream) increased compression ratios
will take time. Sure the spec says 'on the fly', but does it really happen
that way? No! I don't see anyone attempting to do data compression on T1
lines for exactly that reason. Your idea of client selective compression
ratios would help that some, but is it enough to warrant the trouble and
expense? Probably not, the rest of the backup industry certainly doesn't
seem to be running in that direction.

I suspect our disagreement comes over the relative importance of media
capacity vs backup window. It's unfortunate but true that hard drive size
has eclipsed backup media size again, so we're stuck with most clients
needing autochangers. Once autochangers are assumed then media size becomes
less critical, and the backup window and backup speed becomes important.
Retrospect does a pretty good job of reducing the backup data set, almost as
good a job as my former favorite - Palindrome.

It may be of interest that several vendors have looked at on the fly data
compression for things like network traffic, but have always abandoned the
idea because of performance issues. 

-----Original Message-----
From: matt barkdull
To: retro-talk
Sent: 2/26/01 2:36 PM
Subject: RE: Transfer Rates - Dantz Help?

>Matt, if you think you're actually getting 4:1 real world compression
out of
>a modem, I suggest you read some of the research on the subject.

I never said that... What I did say:

>Yes, I know that advertised and what you really get are totally
>different, but all I know is that if something is advertised at 4:1,
>it will be more likely to get at least 2:1 than 2:1 is likely to.

If I attempt 4:1 compression and only get 2-3% on some files, but on 
some files I got 2-3:1, what did I lose or gain?

If I attempt 2:1 compression and only get 2-3% on some files, but on 
some other files I got 1.2 - 1.5:1, what is the advantage?

Keep reading before replying...

4:1 compression under V.42bis is an on the fly compression.  Granted 
the data rates under that specification are a lot less.  (up to 1.5Mb 
vs up to 320MB).  (V.42bis was also written for use on T1.)

What I don't understand is why there is not an option for greater 
compression even if the cost is speed.

For example, let's say I have 4 computers holding a total of 8GB to 
backup over the weekend and my tape drive is only 4GB without 
compression.  With 2:1 compression I am  guaranteed it will not fit. 
With 4:1 compression it might.  It all depends on the data.  I really 
don't care if it takes 24 hours to do it, because the backup goes 
over the weekend.  It should be an option.


BTW - if you backup JPEGs, GIFs, MPEGs, and several others you will 
not compress these files at all. In fact, any compression done on 
them usually results in a larger file.  These are compressed already 
with a compression scheme that is far better than anything else for 
the specialized purpose they do.  Yes, I know about the data losses 
with certain compression schemes, like JPEG will discard information. 
That is totally unacceptable with backup systems.

If you are backing up a normal desktop system, you should be getting
about
1.5:1 compression on average.  If you have systems with a lot of 
graphic files or already compressed files, then yes, the compression 
will be closer to 1:1 (non-compression).


The point is, that you try to achieve 200% and only get 150%, where 
as if you try for 400% and get 200%, you've done better.  If the 
sacrifice was speed, then have a checkbox list that says:

   O  Software Compression 4:1
   O  Software Compression 2:1
   O  Hardware Compression

And let the users decide.





--
----------------------------------------------------------
To subscribe:    [EMAIL PROTECTED]
To unsubscribe:  [EMAIL PROTECTED]
Archives:        <http://list.working-dogs.com/lists/retro-talk/>
Search:  <http://www.mail-archive.com/retro-talk%40latchkey.com/>

For urgent issues, please contact Dantz technical support directly at
[EMAIL PROTECTED] or 925.253.3050.


--
----------------------------------------------------------
To subscribe:    [EMAIL PROTECTED]
To unsubscribe:  [EMAIL PROTECTED]
Archives:        <http://list.working-dogs.com/lists/retro-talk/>
Search:  <http://www.mail-archive.com/retro-talk%40latchkey.com/>

For urgent issues, please contact Dantz technical support directly at
[EMAIL PROTECTED] or 925.253.3050.

Reply via email to