Date: Tue, 26 Jan 1999 12:27:53 +1030 From: Greg Lehey <grog@lemis.com> To: "Jasper O'Malley" <jooji@webnology.com> Cc: John Saunders <john.saunders@nlc.net.au>, freebsd-isp@FreeBSD.ORG Subject: Re: ccd and vinum Message-ID: <19990126122753.B66239@freebie.lemis.com> In-Reply-To: <Pine.LNX.4.02.9901250835130.19608-100000@mercury.webnology.com>; from Jasper O'Malley on Mon, Jan 25, 1999 at 08:56:35AM -0600 References: <19990125131628.C36690@freebie.lemis.com> <Pine.LNX.4.02.9901250835130.19608-100000@mercury.webnology.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Monday, 25 January 1999 at 8:56:35 -0600, Jasper O'Malley wrote:
> On Mon, 25 Jan 1999, Greg Lehey wrote:
>
>> I think it might be possibly better to say that you combine PPs to
>> make LPs, and LPs to make LVs.
>
> Sort of, but not quite. LVs are made up of LPs. But exactly how many PPs
> are in each LP of a particular LV depends on the "mirror" property of an
> LV. If it's not mirrored, there is exactly one PP per LP. If it's a
> simple mirror, there are exactly two PPs for each LP in the LV. If it's a
> double mirror, there are exactly three PPs for each LP in the LV. Each LP
> in a particular LV has the same number of PPs as every other LP in that
> LV. That's where the jump from "physical" to "logical" is made.
Ah. What you're saying is that each LP also has the same storage size
as a PP (4 MB), and that you're combining a number of them to make up
the size of the LV?
>> In this case, a PP corresponds to a subdisk (but it's less
>> flexible), an LP corresponds to a plex, and an LV corresponds to a
>> volume.
>
> I think I'm beginning to see the difference...in the LVM, PPs are used are
> used as simple building blocks, all of a uniform size. I can make a LV of
> any size I want (as long as it's a multiple of the PP size) by picking off
> PPs from anywhere within the VG. That's also how I can grow the
> filesystem, by simply adding more PPs (via LPs) to the LV.
>
> The only thing that's got me confused about Vinum is whether or not a plex
> has to exist within a single subdisk, or whether it can span subdisks,
OK. We're looking at this from a different side. If I've understood
you correctly, a 16 MB mirrored LV would look like:
-------------------------
| | |
| PP0 | PP1 | LP 0 (0-3 MB)
| | |
------------------------
| | |
| PP2 | PP3 | LP 1 (4-7 MB)
| | |
------------------------
| | |
| PP4 | PP5 | LP 2 (8-11 MB)
| | |
------------------------
| | |
| PP6 | PP7 | LP 3 (12-16 MB)
| | |
------------------------
In Vinum, a 16 MB mirrored volume made of 4 MB subdisks might look
like this:
Plex 0 Plex 1
-------------------------
| | |
| SD0 | SD1 | (0-3 MB)
| | |
------------|-----------
| | |
| SD2 | SD3 | (4-7 MB)
| | |
------------|-----------
| | |
| SD4 | SD5 | (8-11 MB)
| | |
------------|-----------
| | |
| SD6 | SD7 | (12-16 MB)
| | |
------------------------
The result is the same, but the concept of plexes is pretty much the
opposite of the concept of LPs. In particular, I don't see how you
can stripe effectively with the LVM approach.
This diagram should also answer your question above: plexes contain
one or more subdisks.
> and whether or not two plexes of a particular volume must take up
> the same number of subdisks (or have roughly the same subdisk
> layout, at least).
No. There's no constraint there.
Greg
--
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-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990126122753.B66239>