HP3000-L Archives

August 1996, Week 3

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Mark Klein <[log in to unmask]>
Reply To:
Mark Klein <[log in to unmask]>
Date:
Tue, 20 Aug 1996 09:55:32 PDT
Content-Type:
text/plain
Parts/Attachments:
text/plain (53 lines)
Goetz wrote:
 
>Jeff Woods wrote:
>>
>> In general, the recommendation from most drive and backup vendors has
>> usually been to not use compression in both modes.  I would suggest trying
>> compress in the drive without the software compressing, in the software
>> without the drive compressing, and also with compression happening in both
>> places.  Whichever is best on your system with your data and your software
>> and your tapedrive is best for you.  Other environments may find other
>> solutions better.
 
>I second Jeff here, if you have the option of both HW and SW compression
>you should try it out. Depending on the CPU speed of your system, SW
>compression might get you a shorter turnaround time for your backup
>(less chance that the IO bandwidth of the bus or the device has the CPU
>idle), OTOH if you do ONLINE backup you may want to avoid the extra CPU
>power needed for SW compression.
 
Actually, there are cases where multiple levels of compression are
usefull. But first, let me describe the various types of compression
available.
 
Hardware compression is typically LZW or some variant thereof. This type
of compression uses a dictionary of repeating strings that can be
dynamically determined on the fly and as such, doesn't require the
dictionary to be stored with the data as with other types of
compression. There are other types of hardware compression available,
but LZW is the most common found on compressing tape drives. LZW can
also be done in software.
 
Since the compressibility of the data really depends on the data itself,
there are instances where negative compression will be achieved as well
as instances where very large files can compress down to almost nothing.
In fact, I've seen an instance where a large (multi GB) database that
was mostly empty got compressed into < 32K using LZW.
 
Where LZW is not effective is in trying to compress something already
compressed with LZW. This can result in negative compression (the
resulting data actually gets larger). For that reason, I wouldn't
recommend using LZW software compression on top of LZW hardware
compression.
 
Another type of compression is called run length compression. This is in
essence a combination of a length tag and a string. The length indicates
how many times to repeat the following string. For example, a line of 80
blanks would be represented by (80," ").
 
Now, using a combination of RLC and LZW one can achieve better levels of
compression than with one or the other method. So, if you want to use
software compression with a hardware compressing tape drive, I would
recommend using RLC compression in software.

ATOM RSS1 RSS2