[linux-lvm] Dumb question

[linux-lvm] Dumb question

[Benoit SERRA]

Is LVM faster when you don't compile it as a module ?

Re: [linux-lvm] Dumb question

[Patrick Caulfield]

On Tue, Jan 23, 2001 at 01:06:03PM +0100, Benoit SERRA wrote:
> Is LVM faster when you don't compile it as a module ?

No.

patrick

Re: [linux-lvm] Dumb question

[Heinz J. Mauelshagen]

On Tue, Jan 23, 2001 at 01:06:03PM +0100, Benoit SERRA wrote:
> Is LVM faster when you don't compile it as a module ?

As Patrick pointed out: no.

You can't almost measure LVM's impact on performance at all.
A different thing is the overhead implied when you take snapshots of logical
volumes because there's additional writes for every write to the original
logical volume necessary which degrades performance almost linear.

> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] New dumb question

[Benoit SERRA]

I build it into the kernel with the 2940 driver, reiserfs and the
software raid (I don't like to have critical drivers as modules)

Stripped LV and NonStripped LVs reported the same bonnie++ result. Is it
normal ?

I'm using 3 2.2 Gb SCSI drives on a 2940. One is for the system, the 2
others are for the LVM. I let the default stripe size, and the default
PE size.

Re: [linux-lvm] New dumb question

[Patrick Caulfield]

On Tue, Jan 23, 2001 at 04:37:37PM +0100, Benoit SERRA wrote:
> I build it into the kernel with the 2940 driver, reiserfs and the
> software raid (I don't like to have critical drivers as modules)
> 
> Stripped LV and NonStripped LVs reported the same bonnie++ result. Is it
> normal ?

If you're doing striped LVs on top of software RAID I'm not really surprised.
Pick one method of striping (LVM *or* RAID) and stick to it but don't mix them 
or who knows where your data will end up!


patrick

Re: [linux-lvm] New dumb question

[Benoit SERRA]

Patrick Caulfield wrote:

> If you're doing striped LVs on top of software RAID I'm not really surprised.
> Pick one method of striping (LVM *or* RAID) and stick to it but don't mix them
> or who knows where your data will end up!
> 
> patrick

Software RAID is build in, but is not used. Nevermind !
I'm not using LVM on top of SOFTWARE RAID for the moment.
If I use LVM on top of SOFTWARE RAID, it will be for redudancy, not for
striping.

My kernel is a 2.4.0 patched for the reiserfs and online ext2 resizing.

Re: [linux-lvm] New dumb question

[Andreas Dilger]

Beniot Sera writes:
> Patrick Caulfield wrote:
> > If you're doing striped LVs on top of software RAID I'm not really
> > surprised.  Pick one method of striping (LVM *or* RAID) and stick to
> > it but don't mix them or who knows where your data will end up!
> 
> Software RAID is build in, but is not used. Nevermind !
> I'm not using LVM on top of SOFTWARE RAID for the moment.
> If I use LVM on top of SOFTWARE RAID, it will be for redudancy,
> not for striping.

It isn't fatal to layer LVM striping on top of software RAID 5, but
I don't think you would get better performance because of too much
overhead from software RAID 5.

You could LVM stripe hardware RAID 5 as long as the LVM stripe size ==
RAID chunk size.  For example, if you have a RAID array with 5 disks
(4 data + 1 parity) and it is using 4kB chunks, then the LVM stripe
size should be n*16kB so that you will write out a full RAID chunk at
one time from LVM.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: [linux-lvm] New dumb question

[Andreas Dilger]

Benoit SERRA wrote:
> I build it into the kernel with the 2940 driver, reiserfs and the
> software raid (I don't like to have critical drivers as modules)
> 
> Stripped LV and NonStripped LVs reported the same bonnie++ result. Is it
> normal ?

It would be a help to LVM users if you could do some simple performance
testing with different LVM stipe sizes.  With larger stripe sizes (64kB
or so) you will likely get better sequential read/write performance, but I
don't think the stripe size will affect random read/write performance.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: [linux-lvm] New dumb question

[Benoit SERRA]

> 
> It would be a help to LVM users if you could do some simple performance
> testing with different LVM stipe sizes.  With larger stripe sizes (64kB
> or so) you will likely get better sequential read/write performance, but I
> don't think the stripe size will affect random read/write performance.
> 

Is there a ratio between stripe size and PE size ?