Hello!
I've got problem with performance on two node stripe configuration. I'm
testing different configurations with script:
#!/bin/bash
x=katalog
for y in a b c d e f g h i j k l m n o p q r s t u v w x y z ; do
echo " $y"
for z in a b c d e f g h i j k l m n o p q r s t u v w x y z ;
do
echo " $z"
mkdir -p $x/$x$y/$x$y$z
for ((i=0;i<=32;i++)) ; do
dd if=/dev/zero of=$x/$x$y/$x$y$z/$i bs=1k count=1 > /dev/null 2>&1
done
sync
done
done
On one node configuration:
volume posix
type storage/posix
option directory /home/export
end-volume
volume brick
type features/posix-locks
subvolumes posix
end-volume
volume server
type protocol/server
option transport-type tcp/server
subvolumes brick
option auth.ip.brick.allow *
end-volume
time of running is about 1m2.149s. Find over directory last for about
45s and removing it - 25s. When I configure client on one node to use
stripe:
volume localhost1
type protocol/client
option transport-type tcp/client
option remote-host 127.0.0.1
option remote-subvolume brick
end-volume
volume remote1
type protocol/client
option transport-type tcp/client
option remote-host 10.240.0.12
option remote-subvolume brick
end-volume
volume stripe
type cluster/stripe
option block-size *:128kB
subvolumes localhost1 remote1
end-volume
I'm getting time of create about 2 minutes, find last for 2m10s and
removing directory - 40s.
Where is problem? Connection to remote1 can speed up to 600Mbps over
scp. I'm using glusterfs 1.3.12 and native kernel fuse (kerel version
2.6.27.2).
Regards.
--
rash at konto pl
