Veritas-bu

Re: [Veritas-bu] Max Fragment Size for Disk and LTO4 Based Storage Units

2008-01-31 19:44:32
Subject: Re: [Veritas-bu] Max Fragment Size for Disk and LTO4 Based Storage Units
From: "Ed Wilts" <ewilts AT ewilts DOT org>
To: "JAJA (Jamie Jamison)" <jamisonj AT zgi DOT com>
Date: Thu, 31 Jan 2008 18:33:00 -0600
On Jan 31, 2008 6:21 PM, JAJA (Jamie Jamison) <jamisonj AT zgi DOT com> wrote:
So I just got a SpectraLogic T950 tape library, and after some initial teething pains it's up and running and it's wicked fast, but being an impatient little monkey of a NetBackup administrator I'm wondering if I could make it wicked faster.
 
Faster for backups or faster for restores?  There's a difference and you need to decide which is more important.  Individual file restores or full volume restores?
 
My primary backups go to DataDomain restorers and then are duplicated to tape for offsite storage with Iron Mountain. When I set all of this up I used 2048 megabytes as the maximum fragment size in both the disk based and tape based storage units. This was based on my predecessor's experience with what was the best trade-off between backup speed and file restoration speed with LTO Gen 2 drives. But I'm wondering if I can improve backup performance even more by increasing the fragment size for my disk and LTO4 based storage units without degrading my current file restoration performance. I thought I would send this to the list to find out what people were using as the frag sizes on LTO4 and on disk based storage units such as the DataDomains.
 
Also if I change the fragment size on my tape based storage units how does NetBackup handle tapes that have backups written to them with different fragment sizes. Can this cause problems? Any feedback will be greatly appreciated.
 
You can change the fragment size on the fly and it won't affect recovery of the existing backups.
 
Larger fragments will be better for full-volume restores but slower for individual file restores.  By how much is up to you to measure.
 
We typically use 20GB fragments on LTO-3 destaging from DSSUs and have found that we don't have much of a significant delay in individual file restores.

   .../Ed

--
Ed Wilts, Mounds View, MN, USA
mailto:ewilts AT ewilts DOT org
_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
<Prev in Thread] Current Thread [Next in Thread>