Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 6 May 2001 12:52:20 +0930
From:      Greg Lehey <grog@lemis.com>
To:        "Andrew C. Hornback" <hornback@wireco.net>
Cc:        FreeBSD Questions <questions@FreeBSD.ORG>
Subject:   Re: Raid
Message-ID:  <20010506125219.E39554@wantadilla.lemis.com>
In-Reply-To: <001f01c0d5d8$3ed3e420$0e00000a@tomcat>; from hornback@wireco.net on Sat, May 05, 2001 at 10:57:11PM -0400
References:  <20010506101618.B39554@wantadilla.lemis.com> <001f01c0d5d8$3ed3e420$0e00000a@tomcat>

next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday,  5 May 2001 at 22:57:11 -0400, Andrew C. Hornback wrote:
> On  Saturday, May 05, 2001 8:46 PM, Greg Lehey wrote:
>> On  Saturday, May 05, 2001 12:19 PM, Steve Blanzy wrote:
>>> On Saturday,  5 May 2001 at 12:34:23 -0400, Andrew C. Hornback wrote:
>>>> What RAID if any does free BSD support in a dual SCSI hard
>> drive configuration?
>>>
>> a> (ACK!  Please don't send HTML-ized e-mail to the list!)
>>
>> In fact, his message was multipart-alternative.  Your MUA shouldn't
>> have any difficulty with that.  But please adjust it to make text
>> lines of < 80 characters.
>
> It should have been set to break lines at 76 characters... of
> course, it's a Microsoft client... *shakes his head* Guess I should
> apologize, eh?

Well, not for that.  I was talking to Steve.

>>>> RAID 1 is pure mirroring, which offers the best redundancy but may
>>>> cause a severe penalty on disk subsystem performance.
>>
>> No, RAID-1 gives you the best performance of any RAID setup. The
>> reason why you need at least 3 disks for RAID-5 is because it is
>> slower, and though it would theoretically work with only two disks,
>> it has no advantages over RAID-1 in this configuration.
>
> 	I'm wondering about the explanation here, as I've always been
> told (and taught) that RAID 5 offers the best performance of any
> RAID level.

Don't believe everything you hear.  This one is just plain not true.

> This may be dependant on a few things (i.e. burst transfer rates of
> the drives and the controller), etc.

Nope, that has no influence.

> RAID 1 causes a performance hit during write, as the data to be
> written has to be done twice.

Correct, but that can overlap.

> RAID 1 is really recommended when you absolutely must have the data
> available, because it allows for two copies to be stored.

RAID-5 is almost as reliable.

> 	RAID 5, I would imagine, suffers a performance penalty in the
> whole comparison with the parity record, but, as I mentioned before,
> would perform better if the total burst transfer rate of the drives
> was equal to the burst transfer rate of the controller, if the
> controller's hardware handled the parity.  (My knowledge of exactly
> how a RAID controller works is spotty, at best...)

No RAID-5 implementation that I know compares parity on reading; read
performance is comparable to RAID-1.  But on write, you need:

 - get (read) old contents of block
 - get (read) old contents of parity block
 - compute parity
 - write parity block
 - write data block

This take much longer than any other RAID implementation, with the
notable exception of RAID-4, which does the same operations but
overloads its parity disk.  Take a look at the section "Performance
considerations" in vinum(8) for some details.

Greg
--
When replying to this message, please copy the original recipients.
If you don't, I may ignore the reply.
For more information, see http://www.lemis.com/questions.html
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010506125219.E39554>