[Gluster-users] Performance problem on 2-node stripe.

[Rash]

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