Posted by:
Adrian Phillips
Date:
02-24-2010,
10:00:AM
|
<!--[if gte mso 9]><![endif]--><!--[if gte mso 9]><![endif]-->
I’ve used dedupe on both NFS and CIFS (CIFS currently),
and seen various results – up to %60 savings on Engineering data, and as
low as %20 on CIFS art data (i.e. jpegs). While I like dedupe (as it is
free…), it isn’t as ideal a disk space saver due to it running on a
daily schedule (could increase, but that would effect CPU). Had some pretty good success with a product called Storwize,
which does on the fly compression (saving around %50). Benefit of it is
that it does compression before writes occur, so less writes and see the
savings right away – before it hits the disk. Saves on backup tapes
and time as well, as the data doesn’t have to decompress on read/write to
tape. When it does iSCSI as well, I’ll definitely take a closer look
at it for the environment I’m currently in.
Again, I like dedupe, but it isn’t the be all and end all
I don’t think. Also had issues with various incremental ndmpcopies
when a dedupe is run in between them…
Cheers, -Adrian
From:
studiosysadmins-discuss-bounces@mailman.studiosysadmins.com
[mailto:studiosysadmins-discuss-bounces@mailman.studiosysadmins.com] On
Behalf Of Stefan Maletych
Sent: Wednesday, February 24, 2010 6:21 AM
To: Stefan Maletych
Cc: discuss@studiosysadmins.com
Subject: Re: Data deduplication Sorry, I should specify that a reboot of the server seems to
fix it...although only temporary... Yeah, we've got it running here on 2008 (just
recently) Lately our server has been quite slow with Vista/Win7 and OS
X machines (almost overnight - which makes me suspect sis) The server is fine with Linux, XP32 and XP64, but is abysmal
on the Vista/7 platform along with Leopard and Snow leopard (which rules out
remote differential compression and autotuning...although I did turn them off
and tried all sorts of combinations) When we copy to or from the server, we'll randomly get
errors like "file or folder not found" "cannot copy, file was not
found, etc." when the file and folder do indeed exist! And, these are even files or folders that have been
deduplicated - they're the only copy on the server All updates are installed on the workstations ad server Also, a reboot seems to fix it....it's so strange...
Sent from my iPhone We use SIS on Storage Server 2008.
There is a service that runs periodically (called the Groveler) which
searches for identical files, and replaces references to them with a 'Junction
Point' (the Windows equivalent of a symlink). This is handled at
the filesystem level and is transparent to machines accessing the files across
a share.
If either of the files is changed independantly, a new copy of the file is
created (using copy-on-write) and the junction point removed.
>From my experience, it saved us around 10-15% on our primary storage, and the
load from the groveller service is negligible (it runs as a background priority
thread).
I've not found a gui to control it - all configuration is done via the
SISADMIN command line tool. It also only works at the drive level - you
can't independantly turn it on for folders.
HTH
B.
On 24/02/2010 06:17, Stefan Maletych wrote: Yes, I've read a bit about the netapp stuff - but what I'm
more concerned with are all of your experiences, and recommendations....your
trials and your errors....
_ StudioSysAdmins-Discuss mailing list a
href="mailto:StudioSysAdmins-Discuss@mailman.studiosysadmins.com"StudioSysAdmins-Discuss@mailman.studiosysadmins.com http://mailman.studiosysadmins.com/mailman/listinfo/studiosysadmins-discuss
|
