okay, I figured something out:
if i mount file based btrfs -
dd if=/dev/zero of=/mnt/btrfs.img bs=1024 count=20000000 && cd /mnt
&& mkfs.btrfs btrfs.img
-like so:
mount btrfs.img /btrfs -o compress=lzo,loop=/dev/loop0
compressing changes from a 11:1 to about 11:6 from the nfs client, any kernel!
my testing was originally using loop0 and file based filesystem (if
this make sense?)
gary
On Wed, Jul 31, 2013 at 11:13 AM, Gary Artim <gartim(a)gmail.com> wrote:
question: why does my compression rate change based on the *clients*
kernel (i think)?
situation: testing using btrfs with lzo, when I run with equal kernels
on the server and client I get the same compression (on a 11GB:1.2GB).
When I run on a client with a 2.6.x kernel I get (11GB:6GB), same
file, all nfs4. btw, I dont get the same speed, 11GB file, server to
server, 16secs on btrfs/lzo; 11GB file, client to server 2mins on
btrfs/lzo over NFS.
has anyone tried or can explain?
server: 3.9.9-201.fc18.x86_64
client (in question): 2.6.31.6-2.caos
client (get same compression): 3.9.9-201.fc18.x86_64 or 3.9.9-302.fc19.x86_64
thanks for any enlightening thoughts!