Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 13 Mar 1999 14:29:44 +1030
From:      Greg Lehey <grog@lemis.com>
To:        Ludwig Pummer <ludwigp@bigfoot.com>, Laurence Berland <stuyman@confusion.net>
Cc:        freebsd-questions@FreeBSD.ORG
Subject:   Re: vinum (how to use after creation)
Message-ID:  <19990313142944.X429@lemis.com>
In-Reply-To: <4.1.19990307205103.00a42670@mail-r>; from Ludwig Pummer on Sun, Mar 07, 1999 at 09:02:00PM -0800
References:  <199903071254.MAA02531@franklin.matlink> <4.1.19990307101720.00c13100@mail-r> <19990308143012.M490@lemis.com> <36E34FBD.CB05405A@confusion.net> <4.1.19990307205103.00a42670@mail-r>

next in thread | previous in thread | raw e-mail | index | archive | help
[Format recovered at freebie.lemis.com]

On Sunday,  7 March 1999 at 21:02:00 -0800, Ludwig Pummer wrote:
> You didn't need to send your message twice.
>
> At 08:19 PM 3/7/99 , Laurence Berland wrote:
>> Can someone get a little bit more in depth on the different kinds of plexes?
>> I'm not
>> quite clear what the differences are.
>
> I'll present simplified examples, but they should be enough for you to get
> the general idea.
>
>> concatenated plexes use the complete address space of each subdisk
>> in turn.
>
> That means that the first subdisk has the first X megabytes of the plex,
> the next subdisk has the next Y megabytes, etc. up to the total size of the
> plex.
>
>> A striped plex conforms to RAID 0. The address space is taken from
>> each subdisk in turn in stripes of a specified size. This makes for
>> more even loading in many cases.
>
> If you've got 2 subdisks, then 1 subdisk has half of the plex, and the
> other had the other half. The data is "interleaved," meaning that that
> instead of having the first half of the plex on one drive, every other
> megabyte is on the first drive, and the 'other' every other megabyte is on
> the second drive.

More or less correct.  Instead of "megabyte" say "stripe".  You can
decide how big to make the stripe, but currently we can't see much
reason not to accept a default of, say, 512 kB.

>> A RAID 5 plex incorporates error recovery: if each subdisk is
>> located on a different physical drive, the plex can continue
>> operation even if any single drive involved in the plex fails.
>
> IIRC, RAID 5 involves striping + parity. 

Correct.

> For a 3 subdisk configuration, data is striped between the first two
> subdisks. The subdisk drive holds parity information for every bit
> contained on the first two subdisks. A '1' bit on the third subdisk
> means that the corresponding bits on the first two subdisks are the
> same, and a '0' means that they are different. So, if any one
> subdisk goes down, it can later be recreated from the other two.

This is, in fact, RAID-4.  RAID-5 differs in that the parity subdisk
changes from one stripe to the next: with 3 disks (the minimum, and
really not enough), the parity block for the first stripe is on the
third subdisk.  The parity block for the second stripe is on the
second subdisk.  The parity block for the third stripe is on the first
subdisk.  The parity block for the fourth stripe is on the third
subdisk, etc.

Greg
--
When replying to this message, please copy the original recipients.
For more information, see http://www.lemis.com/questions.html
See complete headers for address, home page and phone numbers
finger grog@lemis.com for PGP public key


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?19990313142944.X429>