[Gluster-users] GlusterFS in comparison with NFSv4

[Klemens Kittan]

Hello,

For our cluster I've tried different Cluster File System. GlusterFS I
noticed positively by its simple configuration.


This is my setup:

GlusterFS Server: Dell PowerEdge R630 + MD1200
Debian:	8.4
GlusterFS: 3.5.2
Network: Infiniband

Create the volume (Distributed Volume):
gluster volume create home-vol transport tcp,rdma \
gluster01.ib.cluster:/srv/glusterfs/home/brick1/brick

On the same setup I have previously placed an NFSv4 server and made
various measurements. In comparison, GlusterFS performs very poorly. Is
it possible with GlusterFS to reach the bandwidth of NFSv4?


These are my measurements:

NFSv4: 
dd if=/dev/zero of=/mnt/gfs/home/benchmark bs=1G count=1 oflag=direct
706 MB/s

dbench -D /mnt/gfs/home -t 1800 10
Throughput 241.034 MB/sec 10 clients 10 procs max_latency=25.370 ms

GlusterFS:
dd if=/dev/zero of=/mnt/gfs/home/benchmark bs=1G count=1 oflag=direct
356 MB/s

dbench -D /mnt/gfs/home -t 1800 10
Throughput 45.1128 MB/sec 10 clients 10 procs max_latency=50.154 ms


Thank you very much,
Klemens

-- 
Klemens Kittan
Systemadministrator

University of Potsdam
Institute for Computer Science
August-Bebel-Str. 89
14482 Potsdam

Tel.    : +49-331-9773125
Fax.    : +49-331-9773122
Email   : klemens.kittan at cs.uni-potsdam.de
XMPP    : kittan at jabber.ccc.de

gpg --recv-keys --keyserver pgp.mit.edu 6EA09333
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 4687 bytes
Desc: not available
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160531/4e7d3fc2/attachment.bin>