[mythtv-users] Raid 5 Performance Tweaking

[Dean Wilson]

Sorry -- the last email should read "defined in the LVM howto
(http://www.die.net/doc/linux/HOWTO/LVM-HOWTO-4.html) as 'chunks of
data'"

On 6/5/06, Dean Wilson <dean.k.wilson at gmail.com> wrote:
> Sorry if this is a stupid (perhaps related?) question:  I have LVM
> installed on top of my raid array.  Should my physical extents
> (defined in the  as "chunks of data") be the same size as the raid
> chunks?  (vgdisplay /dev/VolGroup00 returns a PE size of 32MB.)
>
>     --- Volume group ---
>     VG Name               VolGroup00
>     System ID
>     Format                lvm2
>     Metadata Areas        1
>     Metadata Sequence No  6
>     VG Access             read/write
>     VG Status             resizable
>     MAX LV                0
>     Cur LV                5
>     Open LV               5
>     Max PV                0
>     Cur PV                1
>     Act PV                1
>     VG Size               931.53 GB
>     PE Size               32.00 MB
>     Total PE              29809
>     Alloc PE / Size       26935 / 841.72 GB
>     Free  PE / Size       2874 / 89.81 GB
>     VG UUID               qeWJuY-j7MI-6R6n-UETw-5LyN-knCS-5Um8Ay
>
> Thanks!
>
> On 6/5/06, Dean Wilson <dean.k.wilson at gmail.com> wrote:
> > I recently installed Fedora Core 5 from scratch on a raid5 array of 5
> > disks.  But I didn't see any options for how to configure the raid
> > array, and now I'm wondering if it could be tweaked better for large
> > files (i.e. for use with myth)  I read in the wiki
> > (http://www.mythtv.org/wiki/index.php/RAID, and
> > http://www.tldp.org/HOWTO/Software-RAID-HOWTO-9.html) that a 32K chunk
> > size and 4K block size was the best setup.  However, I notice that
> > when I run 'cat /proc/mdstat' I get the following:
> >
> >      Personalities : [raid5] [raid4]
> >      md0 : active raid5 sdd1[4] sdc1[3] sdb1[2] sda1[1] hda3[0]
> >            976783360 blocks level 5, 256k chunk, algorithm 2 [5/5] [UUUUU]
> >
> > Apparently the default setup for FC5 was 256k chunks.  (Which doesn't
> > state block size -- anyone know how to find that?)  I noticed that the
> > raid performance page doesn't test beyond 32K chunk sizes -- does
> > anyone know which is best?  (I would think that larger chunk sizes are
> > better for large files, but I don't know...)
> >
> > Also, I found the Raid Howto
> > (http://www.tldp.org/HOWTO/Software-RAID-HOWTO-5.html) which states
> > that "RAID-{4,5} performance is severely influenced by" stride
> > settings for mke2fs.  Since I'm using xfs, is there a similar setting
> > I should use?
> >
> > (I guess I should also ask if it's possible to change these settings
> > at this point?  If not, would the performance gain be worth
> > reinstalling?)
> >
> > Thanks in advance,
> > ~Dean
> >
>