[Gluster-users] performance / IOP

[Claudio Kuenzler]

No official document here, but just a quick and dirty dd on my first
glusterfs install (back in February).
Setup: Two servers, replica=2, Hardware Raid 1, SATA Nearline 10k

On a clean system (back in February) without anything else running:

time { dd if=/dev/zero of=/mnt/vmdata/file.tmp bs=32768 count=32768; sync; }
32768+0 records in
32768+0 records out
1073741824 bytes (1.1 GB) copied, 17.3732 s, 61.8 MB/s

real    0m17.594s
user    0m0.028s
sys     0m4.697s


Today I just ran another dd, just for you.. The glusterfs mounted volume is
only used for nightly backups so no current traffic on it (but the disks
themselves are having some I/O's). Note this is no high end system, just a
small Proxmox cluster for some vms.

time { dd if=/dev/zero of=/mnt/pve/vol1/dump/file.tmp bs=4096
count=1000000; sync; }
1000000+0 records in
1000000+0 records out
4096000000 bytes (4.1 GB) copied, 162.628 s, 25.2 MB/s

real    2m42.844s
user    0m1.072s
sys     0m30.541s

..........

 time { dd if=/dev/zero of=/mnt/pve/vol1/dump/file.tmp bs=4096
count=10000000; sync; }
10000000+0 records in
10000000+0 records out
40960000000 bytes (41 GB) copied, 1674.28 s, 24.5 MB/s

real    27m54.772s
user    0m9.151s
sys     4m26.453s

Hope these values help to give you a "starting point".


On Fri, Jun 20, 2014 at 1:03 PM, Uwe Kastens <kisteorg at googlemail.com>
wrote:

> Hi there,
>
> Are there any documents available which allows to get an idea about
> performance and IOPs?
>
> If I understand the concept correctly a brick could be a harddisk or a
> subset of a RAID. So if would simply mirror two bricks on two servers the
> limiting factor would be the NIC between?
>
> Maybe someone could give a starting point.
>
> Kind Regards
>
> Kiste
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140620/c64be724/attachment.html>