I think this is a case for automation instead of micro management. I use a
small script for all new node definitions. This script calculates the size of
the existing collocation groups and automatically inserts the new node in the
group with least data. Additionally, if a collocgroup spans more than n tapes,
a new one is created.
Paul
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Hans Christian Riksheim
Sent: Wednesday, November 30, 2011 21:11
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: [ADSM-L] Collocaton by group
On Wed, Nov 30, 2011 at 8:16 PM, Allen S. Rout <asr AT ufl DOT edu> wrote:
> How many tapes do you have? Now envision _all_ of them being mounted
> for any restore, even a relatively small one. That's the end state of
> colloc=no: to a first approximation, every node, every filespace,
> every directory, is smeared across a maximum number of tapes, with no
> tendency to re-group data which is predictably related
Mounting 1000 tapes to restore 20G would not be good. I agree.
> Collocation is about paying, in unused tape space, for efficient
> restores; I'd suggest you view that as a bargain, not a burden.
Having our tapes extremely under utilized to achieve some kind of
acceptable performance is a burden.
I only see using disk as a solution here. At least for incrementals.
We don't have the time to do extensive micro management to reach a
compromise between unacceptable tape utilization and unacceptable
restore times.
Hans Chr.
|