ADSM-L

Re: [ADSM-L] the purpose of "file" device class

2009-09-22 05:11:55
Subject: Re: [ADSM-L] the purpose of "file" device class
From: Michael Green <mishagreen AT GMAIL DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Tue, 22 Sep 2009 12:09:58 +0300
On Tue, Sep 22, 2009 at 11:40 AM, Loon, EJ van - SPLXM
<Eric-van.Loon AT klm DOT com> wrote:
> Hi TSM-ers!
> At this moment we are using a diskpool with a VTS-like (DL4106 by EMC)
> storage pool as nextpool.
> I too am looking at a FILE pool to replace this in the future, just to
> prevent a vendor lock-in for our TSM environment and of course the
> possibility to use de-dup.
> The only problem I see for using large FILE (100 Tb +) pools is the size
> of the filesystems on the host running the TSM server. Now it's AIX, but
> in the future we are likely to migrate to Linux.
> What it the server crashes and you end up with one ore more corrupted
> filesystems on you TSM server. How long will a filesystem check run on
> such large filesystems?

As somebody has already mentioned in order to achieve shorter fsck
times make your setup more granular. I have good experience with XFS
under Linux. Can recommend it to anyone. It's fast, stable and
developed. If you choose ext3 route for your FILE devclass, pay
attention to -T largefile option in mke2fs. It will shrink fsck times
considerably.

--
Warm regards,
Michael Green