| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> OT RE: Fwd: RE: RAID or NOT to RAID?
Ha ha ha!
The problems always appear when people apply theory to the real world.
That's when the principles and the reasoning is consistent throughout and makes sense, ... but it has no relevance to reality.
Beware of silly people who walk around with theoretical models in their heads! They are out of this world. I am sure there is a label for this in the DSM-IV.
Some people even start fights because their "maps" of the world don't match! It's truly hilarious to watch all these silly people. Unfortunately sometimes they hurt innocent bystanders, or they cost other people money in useless implementations.
So keep an eye out for explanations that lull you into feeling comfortable or too sure of yourself...
: )
Patrice Boivin
Systems Analyst (Oracle Certified DBA)
Systems Admin & Operations | Admin. et Exploit. des systèmes Technology Services | Services technologiques Informatics Branch | Direction de l'informatique Maritimes Region, DFO | Région des Maritimes, MPO
E-Mail: boivinp_at_mar.dfo-mpo.gc.ca <mailto:boivinp_at_mar.dfo-mpo.gc.ca>
-----Original Message-----
From: Gaja Krishna Vaidyanatha [SMTP:oraperfman_at_yahoo.com]
Sent: Friday, August 10, 2001 9:41 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Fwd: RE: RAID or NOT to RAID?
All,
In my experience I have found "NO" performance
differences between RAID 0+1 and 1+0. We have done
some very extensive tests in this regard. I have no
idea where the additional 50% write performance loss
for 1+0 is coming from. RAID 0+1 and 1+0 both perform
comparably at the same levels for writes.
Regardless of 1+0 or 0+1, the overhead of multiple
writes (as a result of 2-way mirroring) does not
exceed 15-20% at worst. When the logical volumes are
supported by cache, the write overhead is even less.
Some tests that we have done show 10ms write times for
single drives and when mirrored the response time goes
up to 11.5-12ms at worst.
Let's first try to understand the difference between
the physical characteristics of 0+1 and 1+0. In the
former, you stripe first and you mirror what you
striped. In the latter, you mirror first and then you
stripe over what you mirroed.
The difference between the 2 RAIDs, is in the
"available I/O capacity (IOPS, Transfer Rate)" of the
logical volume, when a drive failure occurs. With RAID
0+1, you will always lose 50% of your logical volume's
I/O capacity (regardless of your stripe width), when
you lose 1 drive.
For e.g, if you have a 2-way striped mirrored volume,
when you lose 1 drive in the volume, with RAID 0+1 you
will lose 50% of the logical volume's I/O capacity, as
the whole member of the logical volume is shot,
because the stripe becomes totally invalid. With a
comparable configuration using RAID 1+0, you will lose
only 25% of your logical volume's I/O capacity.
When the striped volume is a 4-way stripe, again with
1 drive loss, with RAID 0+1 you will lose 50% of I/O
capacity versus with RAID 1+0, you will only lose 12
1/2 %. As your stripe-width increases, the percentage
loss in the I/O capacity for RAID 1+0 decreases.
You should (when possible) consider using RAID 1+0 any
day when compared to RAID 0+1.
Cheers,
Gaja
--- Christopher Spence <cspence_at_FuelSpot.com> wrote:
> There is significant differences due to the fact of
> 1+0 you have 2 raid 1
> (which generally have -15% hit in performance on
> writes but increased
> throughput on reads). Raid 0+1 you have a single
> raid 1 stripe, so your
> write performance suffers 50% less than that of Raid
> 1+0. Which I would say
> (never played with Raid 1+0 a lot) but I would say
> here is a decent
> guesstimate scale.
>
>
> Raid 0+1 15% loss in write performance over same
> 4 spindles
> Raid 1+0 30% loss in write performance over same
> 4 spindles
> Raid 5 60% loss in write performance over same 4
> spindles
>
> Anyone who has extensively played with this, I would
> be interested to see
> exact numbers, but confident these numbers are
> fairly close.
>
> "Do not criticize someone until you walked a mile in
> their shoes, that way
> when you criticize them, you are a mile a way and
> have their shoes."
>
> Christopher R. Spence OCP MCSE MCP A+ RAPTOR CNA
> Oracle DBA
> Phone: (978) 322-5744
> Fax: (707) 885-2275
>
> Fuelspot
> 73 Princeton Street
> North, Chelmsford 01863
>
>
>
>
> -----Original Message-----
> Sent: Thursday, August 09, 2001 12:55 PM
> To: Multiple recipients of list ORACLE-L
>
>
> Thanks for the enlightenment. What is the
> performance difference, from an
> Oracle database standpoint, between RAID 0+1 and
> RAID 1+0?
>
> Unfortunately one cannot change the configuration on
> Government Furnished
> Equipment without a bit of a hassle. In a more
> perfect world, of course,
> but in real life, not likely.
>
>
>
>
> Paul Drake
>
> <paled_at_home.c To:
> Multiple recipients of list
> ORACLE-L
> om>
> <ORACLE-L_at_fatcity.com>
>
> Sent by: cc:
>
> root_at_fatcity. Subject:
> Re: Fwd: RE: RAID or
> NOT to RAID?
> com
>
>
>
>
>
> 08/09/2001
>
> 03:10 AM
>
> Please
>
> respond to
>
> ORACLE-L
>
>
>
>
>
>
>
>
>
> Hi all.
> I've been disconnected for awhile - found myself out
> in Nebraska. Installing
> 81715/Win2000 on a new Dell PowerEdge 2500. Only had
> 6 drives on 1 IO
> channel configured as 3 "containers" - RAID 1. The
> front grille reminds me
> of an electric shaver. It was the kinda trip that
> ended with me pulling a
> hamstring while unplugging a power cord for my
> notebook. Gotta stay hydrated
> and get that daily banana for potassium.
>
>
> repeat after me ...
>
> RAID 10 != RAID 0+1
> RAID 0+1 != RAID 10
>
> Even with only 4 drives, when they might seem the
> same,
> RAID 0+1 is stripe first, then mirror.
> RAID 10 is mirror first, then stripe.
> As someone once said, the best way to tell how its
> configured, is to pull a
> drive out of a hot swap bay, put it back in and see
> how many drives
> re-silver. A corollary would be - pull one drive -
> and then pull another
> non-adjacent drive (e.g. in the other cage). If its
> RAID 01 - you're
> completely hosed. So much for non-destructive
> testing. :)
>
> oh yeah, and "it depends".
>
> If you only need 8KB or 64 KB blocks at one time, go
> for neither, and just
> separate files onto different RAID 1 volumes of 2
> disks each. If you're
> daring, don't even bother to use hardware RAID for
> the online redo logs -
> and just duplex them with multiple log members of a
> redo log group.
>
> After you locate what your point of contention is -
> either move the hot
> spots out to dedicated drives, or add more drives to
> the volume that has the
> most I/O.
>
> If you need massive amounts of data from full table
> scans - go for deeper
> stripes. Even numbers of drives in a volume are
> preferable for RAID 0
> stripes, odd for RAID 3,5. This makes it easy to
> calculate the stripe depths
> as a multple of the db_block_size and OS io_size in
> your head.
>
> Gaja wrote a great section on this topic in the
> Performance Tuning 101 Book.
>
> And you can "fix" the RAID configuration by simply
> deleting the existing
> RAID config - and starting from scratch. I had a
> site where a Dell Tech took
> a perfectly good 4 x RAID 1 (8 drives) config and
> turn it into a single RAID
> 0+1 config. <Ctrl-A> at boot gets you into where you
> can wipe it clean and
> start from scratch - assuming that you can wipe the
> slate clean.
>
> And I'll beat Joe T to the punch - since you're
> going to have to re-install
> the OS - Dell boxes run Linux pretty well. Just that
> Dell still sucks badly
> for calling RedHat Linux "Linux 7" on their store
> website. That still
> pisses me off.
>
> Linux != RedHat.
>
> Paul
>
>
> tday6_at_csc.com wrote:
> >
> > In a previous job I had to deal with this issue.
> WinNT 4.0 on a dual
> > processor Dell box with 24G of RAID. I had
> specified RAID 0 + 1 but
> > "someone" knew better and got it with RAID 5 (5 is
> obviously better
> > than 0). The SA wouldn't or couldn't reconfigure
> and the job needed
> > to get done.
> >
> > It was for a decision support system (basically
> read-only) and it may
> > be that sort of a system is less impacted. I
> abandoned any thought of
>
=== message truncated ===
=====
Gaja Krishna Vaidyanatha
Director, Storage Management Products,
Quest Software, Inc.
Co-author - Oracle Performance Tuning 101
http://www.osborne.com/database_erp/0072131454/0072131454.shtml
__________________________________________________
Do You Yahoo!?
Send instant messages & get email alerts with Yahoo! Messenger.
http://im.yahoo.com/
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Gaja Krishna Vaidyanatha
INET: oraperfman_at_yahoo.com
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing
Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
-- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Boivin, Patrice J INET: BoivinP_at_mar.dfo-mpo.gc.ca Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051 San Diego, California -- Public Internet access / Mailing Lists -------------------------------------------------------------------- To REMOVE yourself from this mailing list, send an E-Mail message to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in the message BODY, include a line containing: UNSUB ORACLE-L (or the name of mailing list you want to be removed from). You may also send the HELP command for other information (like subscribing).Received on Mon Aug 13 2001 - 10:13:08 CDT
 |
 |