Andrew Carlson wrote:
> I know that the recommendation is to use as many physical volumes as
> possible. In the near future, volumes are going to become available to
> me to use as many as 15 volumes per database instance.
>
> Here is what I am wondering. It seems that, if I create a 5GB volume
> per disk, for a total of 75GB capacity, that TSM would tend to write to
> one volume for quite some time if it were JBOD. If I made it striped,
> if I understand striping correctly, it will put the first chunk on the
> first disk, second chunk on the second disk, etc. Then create one large
> DB vol, which will effectively be spread across multiple volumes without
> me having to make sure of the housekeeping.
>
> Any ideas, or real world experience?
>
The TSM database has inherently a tendancy to 'stripe'. That is, after a
few months of use, even on a jbod you'll see all disks used equally often.
Now if you lose one disk in a raid-0, you'll lose the whole FS, thus all
of you db volumes. If you have multiple raid-sets (or jbod) you'll just
loose one db volume, which might make recovery a lot faster (depending
on logmode settings and mirroring).
I have the feeling that raid-0 might even slow TSM performance down when
compared to JBOD. At best, I can't think of any advantage of raid-0.
> For the record, in case anyone asks, I will be at TSM5.3 by the time I
> do this, using a Fastt700 for the database volumes.
>
> Thanks!
--
Met vriendelijke groeten,
Remco Post
SARA - Reken- en Netwerkdiensten http://www.sara.nl
High Performance Computing Tel. +31 20 592 3000 Fax. +31 20 668 3167
"I really didn't foresee the Internet. But then, neither did the
computer industry. Not that that tells us very much of course - the
computer industry didn't even foresee that the century was going to
end." -- Douglas Adams
|