[Gluster-users] high wa% values

[samuel]

Hi folks,

Just starting playing with gluster because I'd like to move from an NFS (v3)
storage system to glusterfs. I downloaded sources from last 3.2 version and
compiled in all the clients and servers (ubuntu 11.04). Just installed
default parameters both for NFS and gluster in order to compare "basic"
installations.

In a kvm virtual environment using the same instance (copying the disk image
and runing both of them) the preformance using NFS is significantly much
better, specially regarding percentage of processes in waiting status.
Absolute values using hdparm or fio are right but gluster's backend virtual
machines stalled from time to time when processes are in waiting status.

Backend filesystem is
1)NFS with RAID 1+0 and XFS,
2) while gluster are just plain SATA XFS formated disks.

I've read gluster has lower performance accessing huge number of small files
compared to NFS but, in theory, the rest of scenarios gluster offers better
performance.
Can anyone point to some documentation where I can improve gluster
behaviour? Or any suggestion|idea to improve the storage system?

Thank you very much in advance,
Samuel.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20110701/f7a48c90/attachment.html>