[Gluster-users] mirror, performance

[Kun Richárd]

Hi, 

I built a mirrored fs with 2 nodes, by the "Automatic File Replication
(Mirror) across Two Storage Servers" example.
I tested it's performance, with big files, it was good enough.
If I tried to extract a kernel-source-tree (with many small files), it was
extremly slow.

I fingered parameters like 
 - performance/writebehind: flush-behind to on, 
 - option thread-count from 8 to 16,
but the result was the same
Extracting kernel-2.6.16.tar.gz takes 23 minutes, without glusterfs the
machine extracted the tarball within 10 sec!

Any suggestions?

Thanks:

--
Üdv,
 Ricsi