[Gluster-devel] write perf vs read perf

[Amar S. Tumballi]

Hi Albert,
 We have 'write-behind' performance translator to enhance write speed while
writing to the remote node. Do use write behind on the client spec file,
(above unify) to get more write performance.

Regards,
Amar

On Dec 7, 2007 10:10 PM, Albert Shih <Albert.Shih at obspm.fr> wrote:

> Hi all
>
> I'm using on my cluster glusterfs with unify with numa scheduler.
>
> When I write into gluster partition on one node, of course it's fast
> (write on local disk)
> after that when I read from another node the perf is good ~ 95Mo/s (on
> 1Gbits/s switch)
>
> But if I write into this file from another node the perf is very bad ~
> 8Mo/s. I known that's not the normal situation, but this can append when
> the file is create on one node, and append in other node sometime later.
>
> This performance ~8Mo/s on 1Gits/s network is normal ? How can I improve
> this ?
>
> All node of my cluster is identical and have 16Go Ram, two disk at
> 15000rpm.
>
> Regards.
>
> --
> Albert SHIH
> Observatoire de Paris Meudon
> SIO batiment 15
> Heure local/Local time:
> Ven 7 déc 2007 17:36:29 CET
>
>
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at nongnu.org
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>



-- 
Amar Tumballi
Gluster/GlusterFS Hacker
[bulde on #gluster/irc.gnu.org]
http://www.zresearch.com - Commoditizing Supercomputing and Superstorage!