← Back to team overview

sony-vaio-z-series team mailing list archive

Re: Major RAID speed improvement (for advanced Linux users)

 

Hi Knuth,

No, in fact that's the same setup: md127 (and then md126) are the default
assignments that are done by mdadm if you don't specify a specific md
device.

Personally I use this in my /etc/mdadm.conf:

ARRAY metadata=imsm UUID=xxxxxxxx:xxxxxxxx:xxxxxxxx:xxxxxxxx
ARRAY /dev/md0 container=xxxxxxxx:xxxxxxxx:xxxxxxxx:xxxxxxxx member=0
UUID=yyyyyyyy:yyyyyyyy:yyyyyyyy:yyyyyyyy

(it took me a whole evening to figure these 2 lines out)

Therefore the MD device for the imsm (Intel Matrix Storage Management)
container interface will be /dev/md127 (because I don't specify a specific
device for it) and the MD device on top of it (the one which will be used
for storage) will be /dev/md0, which is a convenient name imho.

Best regards,

Joël


On Fri, Dec 3, 2010 at 1:31 AM, K. Posern <quickhelp@xxxxxxxxx> wrote:

> Hi Joel,
>
> thanks for your answer. Because in your example you were speaking about
> md0 I thought it was a somewhat different setup.
>
> I really like the mdadm approach that basically unmaps the fakeraid for
> you via the md127 and then gives you a normal disk to play with (md126)
> :) ... I don't like the naming of dmraid.
>
> But as you said it is handled transparently like a (true) linux software
> raid --> hence you can use all the (e.g. monitoring) infrastructure of
> an mdadm linux software raid :)
>
> Knuth
>
> ---
>
> On 12/02/2010 04:37 PM, Joël Bourquard wrote:
> > Hi Knuth !
> >
> > Well actually I believe you have exactly the same setup as me.  And it
> > *is* a linux software raid - it just happens to be setup with the
> > parameters read from the Intel fakeraid metadata that is on the disks.
> > But the operation is fully software.
> >
> > /dev/md126 is the linux software raid, which uses /dev/md127 as the
> > underlying interface to the metadata.  You can see that in /proc/mdstat:
> >
> > md127 : inactive sdd[3](S) sda[2](S) sdb[1](S) sdc[0](S)
> >       9028 blocks super external:imsm
> >
> > But /dev/md126 is really a true linux software raid0, with a 128kB
> > stripe size.
> >
> > Yes, many distros (including Gentoo also) still use initrd's which don't
> > include mdadm 3.0+, which is required to handle the intel metadata.
> > That's why I recommend to copy a static compile of it, to the initrd.
> >
> > Best regards,
> >
> > Joël
> >
> >
> > On Thu, Dec 2, 2010 at 9:31 PM, K. Posern <quickhelp@xxxxxxxxx
> > <mailto:quickhelp@xxxxxxxxx>> wrote:
> >
> >
> >     Hi Joel,
> >
> >     Thanks for the report.
> >
> >     I ca add though:
> >
> >     If you use mdadm instead dmraid to access your NATIVE fakeraid
> (without
> >     a linux software raid) you get actually the same results :)
> >
> >     hdparm -tT /dev/md0
> >
> >     /dev/md126:
> >      Timing cached reads:   9010 MB in  2.00 seconds = 4506.74 MB/sec
> >      Timing buffered disk reads: 1498 MB in  3.00 seconds = 498.81 MB/sec
> >
> >     And a little side remark:
> >
> >     For all I know: mdadm is also favored by intel going forward over
> dmraid
> >     - but ubuntu still favors dmraid by default (I think).
> >
> >     Best,
> >
> >     Knuth
> >
> >     On 12/02/2010 12:42 PM, Joël Bourquard wrote:
> >     > Actually I think buffered reads are the same, it's just some
> >     measurement
> >     > variance that we see here :-)
> >     >
> >     > I noticed a major improvement when launching huge apps (firefox,
> >     > evolution, OpenOffice) for the first time, and of course when
> >     copying files.
> >     >
> >     > Other than that, the general speed of the system was already very
> >     high -
> >     > so it's difficult to rule out the placebo effect. But it's
> >     certainly not
> >     > running any slower than before, far from it! And numbers are
> >     consistent.
> >     >
> >     > Best regards,
> >     >
> >     > Joël
> >     >
> >     > --
> >     > Sent from my Android phone.
> >     > Please excuse my brevity.
> >     >
> >     > On Dec 2, 2010 6:31 PM, "Simon Brown" <lists@xxxxxxxx
> >     <mailto:lists@xxxxxxxx>
> >     > <mailto:lists@xxxxxxxx <mailto:lists@xxxxxxxx>>> wrote:
> >     >
> >     >
> >     >
> >     > _______________________________________________
> >     > Mailing list: https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series>
> >     <https://launchpad.net/%7Esony-vaio-z-series>
> >     > Post to     : sony-vaio-z-series@xxxxxxxxxxxxxxxxxxx
> >     <mailto:sony-vaio-z-series@xxxxxxxxxxxxxxxxxxx>
> >     > Unsubscribe : https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series>
> >     <https://launchpad.net/%7Esony-vaio-z-series>
> >     > More help   : https://help.launchpad.net/ListHelp
> >
> >
> >
> >     _______________________________________________
> >     Mailing list: https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series>
> >     <https://launchpad.net/%7Esony-vaio-z-series>
> >     Post to     : sony-vaio-z-series@xxxxxxxxxxxxxxxxxxx
> >     <mailto:sony-vaio-z-series@xxxxxxxxxxxxxxxxxxx>
> >     Unsubscribe : https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series>
> >     <https://launchpad.net/%7Esony-vaio-z-series>
> >     More help   : https://help.launchpad.net/ListHelp
> >
> >
>
>
>

References