[Gluster-users] mount via nfs vs via gsluterfs

[Fyodor Ustinov]

Hi!

I compare fs performance by bonnie++ between fs mounted via nfs and via 
glusterfs.

Volume Name: test
Type: Distributed-Replicate
Status: Started
Number of Bricks: 2 x 2 = 4
Transport-type: tcp
Bricks:
Brick1: stc1:/mfs
Brick2: stc2:/mfs
Brick3: stc3:/mfs
Brick4: stc4:/mfs
Options Reconfigured:
nfs.trusted-sync: on


for glust-d: mount -t nfs -o mountproto=tcp stc1:/test /mnt
for glust-d2: mount -t glusterfs stc1:/test /mnt

(both test on the same server)

bonnie++ -d /mnt -u root


Result:

1.96,1.96,glust-d,1,1303161226,8G,,1274,96,51734,3,27419,5,4026,97,102569,8,453.2,10,16,,,,,434,2,851,3,291,2,433,2,1674,5,818,4,6886us,1146ms,35367ms,15497us,425ms,88825us,32546us,651ms,5838ms,8118us,4077us,25173us
1.96,1.96,glust-d2,1,1303221333,8G,,2,9,14878,8,10011,5,4634,98,97307,5,271.4,9,16,,,,,447,7,6837,16,789,6,415,4,2035,6,574,6,4074ms,3225ms,3978ms,17431us,49884us,842ms,49997us,9165us,38936us,43988us,8654us,42528us

Decoded table: http://blog.ufm.su/nvsg.html

Dramatically degraded performance in all write operartions.

Why?

WBR,
     Fyodor.