[Veritas-bu] Fragment size settings
2001-10-17 13:04:27
Fragmentation shouldn't be a problem. Not sure about NBU, but other
applications I've worked with won't cut a file in the middle of a fragment,
it will continue to end of current file. So, if you've already backed up
1.5GB and the next file is .7MB, your fragment size will be 2.2GB. Not
using fragmentation will cause problems however, because some systems
(Solaris 2.5 for sure, think they finally fixed in 2.7?) won't allow file
sizes over 2GB in size. So a 10GB tar file would be unmanageable in this
instance.
Ken Zufall
Work: 330.202.4113
Cell: 330.465.8113
Pager: 888.896.3272
E-page: 4403037243.8963272 AT pagenet DOT net
<mailto:4403037243.8963272 AT pagenet DOT net>
-----Original Message-----
From: Rob Worman [mailto:rworman AT yahoo DOT com]
Sent: Wednesday, October 17, 2001 11:56 AM
To: EBU (E-mail)
Subject: Re: [Veritas-bu] Fragment size settings
and I know of one price you pay for the use of fragment
size:
complication of a "non-NBU" tape restore, i.e. reading from
a NetBackup
tape using only tar.
Without fragmentation, a single 10GB backup image = a single
10GB tar
file. But with a fragment size of 2000, that 10GB backup
image is now 5
separate tar files.
Unlike the use of multiplexing, fragmenting is not
necessarily a
"showstopper" when considering this aspect of NetBackup's DR
capabilities, but if you really want to be prepared for
tar-only
NetBackup restores, you MUST understand the effect
fragmentation has on
your media contents. (and you'd be smart to write a script
to automate
all the tar commands...)
HTH
rob
Kyle Adamski wrote:
> The theory behind the fragment size setting on the storage
unit is to
> break the larger tar images into smaller fragments. This
allows
> restores to speed up because in between the fragments is a
tape mark
> (identifying what is on the fragment) that will allow NBU
to scan the
> tape faster for what it is looking for vs reading the
entire image. We
> use them and have them set at 2G....I have never tested
the speed of
> restores with and without them....I have just accepted
them in
> principle and assumed that they worked as advertised. If
you have them
> set at 0(unlimited) there will be no fragmentation.
>
> Hope that helps.
>
> Kyle A. Adamski
>
> kylea AT firstlogic DOT com
>
> 608.788.8154 x3093
>
> -----Original Message-----
> From: Brochart, Fabrice [mailto:BrochF AT europe.stortek DOT
com]
> Sent: Wednesday, October 17, 2001 9:23 AM
> To: 'Sixbury, Dan'; EBU (E-mail)
> Subject: RE: [Veritas-bu] Fragment size settings
>
> Hi ,
>
> I read somewhere that restore is more faster with fragment
size
> (depend on size (depend on backup :) ))
>
> And some customers tells me the same things but I've never
verified.
>
> HTH
>
> F@brice
>
> -----Message d'origine-----
> De : Sixbury, Dan [mailto:sixbury AT celeritas DOT com ]
> Envoyé : mercredi 17 octobre 2001 16:00
> À : EBU (E-mail)
> Objet : [Veritas-bu] Fragment size settings
>
> Has anyone seen any benefits for using fragment size on
the storage units,
> or have reasons to not use fragment sizes? I think the
default is
> unlimited, but I noticed that we had some storage units
that were set to
> 2000 while others had the default of unlimited.
>
> Thanks,
> Dan
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com
_______________________________________________
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>
|
- [Veritas-bu] Fragment size settings, Sixbury, Dan
- [Veritas-bu] Fragment size settings, Brochart, Fabrice
- [Veritas-bu] Fragment size settings, Kyle Adamski
- [Veritas-bu] Fragment size settings,
Zufall, Ken <=
- [Veritas-bu] Fragment size settings, Zufall, Ken
- [Veritas-bu] Fragment size settings, Cornely, David
- [Veritas-bu] Fragment size settings, Jim Horalek
- [Veritas-bu] Fragment size settings, Marelas, Peter
- [Veritas-bu] Fragment size settings, Marelas, Peter
|
|
|