LVM striping on luns

[(Imed Chihi) عماد الشيحي]

Hi,

Since the storage LUNs are RAID 10, then you're already doing striping
at the disk level.

Striping an LVM volume group would make sense if the two LUNs in
question are presented to the host from two different array
controllers.

Typical enterprise storage would have a host access the storage via
two HBAs and two Fibre Channel switches.  Therefore, if the two LUNs
are "seen" from two different controllers (storage processors), then
you'd get the performance benefits of parallel IO on independent paths
with indepent caches.

  -Imed

On 1/27/12, redhat-list-request at redhat.com
<redhat-list-request at redhat.com> wrote:
> Send redhat-list mailing list submissions to
> 	redhat-list at redhat.com
>
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://www.redhat.com/mailman/listinfo/redhat-list
> or, via email, send a message with subject or body 'help' to
> 	redhat-list-request at redhat.com
>
> You can reach the person managing the list at
> 	redhat-list-owner at redhat.com
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of redhat-list digest..."
>
>
> Today's Topics:
>
>    1. LVM striping on luns (Matty Sarro)
>    2. Re: LVM striping on luns (Nicoas C.)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 26 Jan 2012 12:26:06 -0500
> From: Matty Sarro <msarro at gmail.com>
> To: General Red Hat Linux discussion list <redhat-list at redhat.com>
> Subject: LVM striping on luns
> Message-ID:
> 	<CAHUGJcGvgHfayyBKhbPLQrE=7rCGYqNcDfYZQNDPL7wj_aP+Qg at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> Hey everyone.
> I am working with a number of boxes right now. Each one has several LUN's
> from a SAN attached via fibre channel. The arrays on the SAN should all be
> raid 10. Now, what I'm curious about is if I will get any performance boost
> at all from using striping on the logical volumes, or if I should just use
> linear volumes? It seems like writing across each lun with striping makes
> sense, but at the same time it'd be extra CPU load.
>
> Any advice would greatly be appreciated. Thanks!
> -Matty
>
>
> ------------------------------
>
> Message: 2
> Date: Fri, 27 Jan 2012 11:25:23 +0100
> From: "Nicoas C." <rh at nryc.fr>
> To: General Red Hat Linux discussion list <redhat-list at redhat.com>
> Subject: Re: LVM striping on luns
> Message-ID: <4F227B93.3030101 at nryc.fr>
> Content-Type: text/plain; charset=UTF-8; format=flowed
>
> Le 26/01/2012 18:26, Matty Sarro a ?crit :
>> Hey everyone.
>> I am working with a number of boxes right now. Each one has several LUN's
>> from a SAN attached via fibre channel. The arrays on the SAN should all be
>> raid 10. Now, what I'm curious about is if I will get any performance
>> boost
>> at all from using striping on the logical volumes, or if I should just use
>> linear volumes? It seems like writing across each lun with striping makes
>> sense, but at the same time it'd be extra CPU load.
>>
>> Any advice would greatly be appreciated. Thanks!
>
> In a SAN environment you have to consider that one LUN on a RAID can
> influence the performance of the others LUNs on the same RAID as far as
> they share the same disks.
>
> So, on a RAID, if you have two LUNs, "A" being part of a stripped
> filesystem, "B" being independent, if "B" becomes intensively used, it
> will also decrease the performance of "A".
>
> On a stripped filesystem, reads and writes occur on all the Physical
> Volumes of the pool, if one of them ("A" in my example) becomes slow,
> the performance of the whole filesystem will also becomes slow.
>
> Log story short : a single independent LUN can decrease the performance
> of a stripped filesystem.
>
> If you use "linear", you might not really use all the throughput of the
> LUNs at first because your data won't be spread on all of them but if
> one raid becomes intensively used (by a LUN independent of your LVM
> setup) it won't impact the filesystem as much as with stripping.
>
> I hope this helps,
>
> Nicolas
>
>
>
> ------------------------------
>
> __
> redhat-list mailing list
> Unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
> https://www.redhat.com/mailman/listinfo/redhat-list
>
> End of redhat-list Digest, Vol 95, Issue 8
> ******************************************
>

-- 
Sent from my mobile device

Imed Chihi - عماد الشيحي
http://perso.hexabyte.tn/ichihi/