Bacula-users

Re: [Bacula-users] A question about tape performance.

2008-08-22 16:52:02
Subject: Re: [Bacula-users] A question about tape performance.
From: "Erik P. Olsen" <epodata AT gmail DOT com>
To: bacula-users AT lists.sourceforge DOT net
Date: Fri, 22 Aug 2008 22:49:57 +0200
On 15/08/08 23:25, Martin Simmons wrote:
>>>>>> On Fri, 15 Aug 2008 12:14:09 +0200, Erik P Olsen said:
>> With tapes you have the option of using software or hardware compression or 
>> none
>> at all. I have been using software compression several years expecting that 
>> this
>> was the wisest to use.
>>
>> I have now been through a period of rebuilding my system due to a disk 
>> breakdown
>> and that has meant a lot restores of mainly small amount of data. These 
>> restores
>> have all taken a very long time to complete from when the tapes were mounted 
>> to
>> the actual data was restored. I can see that bacula knows from the catalog in
>> which tape file the data may be found. I assume it's very fast to go to the
>> target file using forward space file, but what goes on when it then locates 
>> the
>> data file within the tape file? I believe it's done by a sequential search of
>> the tape file. I can see from the tray monitor that it reads block after 
>> block
>> before the restore takes place. This seems to cost a lot of time especially 
>> if
>> the blocks have to be unzipped. If this is true a better restore performance
>> will be obtained with uncompressed tapes or even with hardware compressed 
>> tapes
>> because here the data is decompressed in flight.
> 
> I don't think this makes any difference.  The software decompression is
> handled by the file daemon, so it works on the backup data, not the blocks.
> There should be no decompression until it reaches the part of the tape file is
> being restored.
> 
> Check that you have Fast Forward Space File = yes in your SD config and that
> it works for your drive.

That I have but I don't know how to check if it works.

> 
>> If all this is true I tend to believe that hardware compression is by far the
>> best method. On the other hand I know that software compression is advocated 
>> by
>> folks more knowledgeable than me.
>>
>> Would someone please explain what's up and down with this sort of 
>> performance?
> 
> I agree with Ryan -- it depends on your tape drive technology.

It also depends how the data is stored on the tape. I'll see if I can find out.

> 
> The other thing to note is that you might not know how full your tapes are
> when using hardware compression.

That I don't understand, AFAIK Bacula continues to write the tape till EOV and
that should be the same with hardware, software, or no compression. Or have I
misunderstood something?

-- 
Erik.

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>