I think Chris' point has more merit than you give credit, Quentin. There are many issues to consider, including segregating the root volume, parity disks, costs of additional shelves/cabinetry/head units, etc. Granted the cost of disk capacity does not scale linearly. However, if you're not using the additional capacity in a volume (30GB database on a volume of 5-6 72GB drives, as in Chris' scenario) - and there may be very good reasons for this - then big drives are definitely not the most cost-effective way to scale capacity. Qtree-level SnapMirror and/or file-level SnapRestore may solve some of these issues related to capacity management, allowing you to put more databases on the same volume. Even still, it would be nice to have an internal 2x9GB volume (maybe even with solid state disks!) to serve as root, rather than "wasting" 72GB drives and shelf slots.
Just musing...
Joe
Joe Luchtenberg Dataline, Inc.
Email: joe.luchtenberg@data-line.com Phone: 757-457-0504 (direct line) 757-858-0600 (front desk) 757-285-1223 (mobile) Fax: 757-858-0606
Please visit us at www.data-line.com
-----Original Message----- From: Quentin Fennessy [mailto:quentin.fennessy@amd.com] Sent: Friday, March 29, 2002 8:29 AM To: Chris Lamb Cc: Melvin, Grant; 'Noll, Kevin'; 'Toasters (E-mail)' Subject: Re: Vol/qtree/dir/file snapshots (was Re: DataONTAP 6.2) ORACLE DATABASES
Chris asked:
Of course, one just begs to ask: WHY is it that we
continue this absurd
notion that bigger is always better? Who the hell needs a
72GB BOOT DISK?
Sure, maybe when Windows(tm) 2004 comes out... :-)
This is a detail that is unnecessary to sweat. Large drives don't have to be full. The cost of capacity is dropping -- at a given time in product life, a 72GB disk is not 8x the cost of a 9GB drive. It may well be 1x the price, this 1/8x the cost!
Leave the disk space empty if its use is not appropriate.
If you need extra spindles for performance, buy them. Don't get caught in the raw capacity trap.
I cannot imagine that it is economical to continue to sell relatively smaller disks, when for the same outlay the disk manufacturers can provide more capacity. The factories can provide X disks (or disk platters) per week. They will optimize appropriately. And they would go out of business if they did not, because the competition will!
As a system administrator we (you and I) all have to exercise judgement on how to best manage these resources. If your boss or your customers override your judgement then something should change. But it won't be the manufacturing economics of disk drives!
Yes, it is a fundamental law of computing that files and
databases will
always expand to fill the available disk space, but why is
it that you
can't even buy a 9GB drive anymore? A 20-30GB database, even if you double or triple it in size, is still going to fit on *one*
shelf with
9- or 18GB drives.
If I had a 30GB database, I'm sure my DBA would agree -
striping 5-6 9's
together makes way, way more sense than throwing piles of
higher-cost 36's
or 72's at it, only to waste massive amounts of space in an
effort to keep
performance up. Of course, our database grows at over 30
million rows per
week, so adding 36's makes sense for us. :-) Still! I
echo the call for
using appropriately-sized drives where it makes sense. :-)
All of my production servers have ridiculously huge boot
drives because
that's the only way Sun (or anyone else) sells them these
days. So I
mirror / & /var, interleave swap, and even after sizing
things to some
absurd degree, there's usually a chunk of 8-10GB of free
space - 'cuz all
the good stuff lives on the filers. Having that local disk
space there
just means that someone is going to want to USE it, and
that's bad - it
means I'd have to actually back up those machines, and I
*hate* backups.
(In the exceedingly rare case where both drives of a
mirrored boot volume
are corrupted or destroyed, it's always going to be faster
to re-Jumpstart
and cfengine the machine from scratch than to restore from
tape anyway.
Local data just messes up all that beautiful automation. :-)
A long time ago I was going to recommend that filers come
with a pair of
internal boot drives - like a mirrored pair of 4GB or 9GB drives _strictly_ for use as the boot volume, with some space for
logs, etc.
There's plenty of room inside the big filer heads for a
pair of drives.
But the move to Flash RAM cards in the newest machines is
even better -
fewer moving parts. :-)
-- Quentin Fennessy Quentin.Fennessy@amd.com Office: 512.602.3873 Cell: 512.694.7489
-
Joe Luchtenberg