Gluster users - Sep 2014 - Issue with tar, ls and rsync on 10G volume

Hi all,

I have a 2 node cluster in on vms in vmware. They are setup in replication
mode.

These servers are just hosting small files for an application. The
application is mainly reading the contents. The volume is 10G in size
(about 2G used) and there are probably somewhere between 60k and 70k files
varying from image and html/css stuff to some application 'blob' type
data.

I seem to be having a few main pain points with using gluster when compared
to nfs.
- on client, using tar to backup directories is painfully slow
- on client, executing ls -R or du is also slow
- on client, rsync is slow

These simple tasks have been roughly taking 2x-3x as long.

(i've just noticed that doing this on the gluster server itself is
significantly faster. I'd rather not need to do it from there, if possible.)

I've tried these changes below to see if they help but only with minimal
improvement.
- on gfs-01/02 nodes, mounted volume with noatime, nodiratime
- on client system, mounted volume with ro
- on client system mounted volume as nfs with ro (i tried adding async, but
option wasn't showing up with 'mount')

These are the specs:
gluster servers (gfs-01/02) - 1 cpu, 2G ram, centos 6.5 64bit, gluster
3.4.2-1 rpms
 memory usage between 50-60% and cpu < 5%

Is there anything that can be done to improve this? I would appreciate any
suggestions or feedback.

Thanks,

Sijis
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140916/7fe41ebf/attachment.html>