[Gluster-users] distributed vs unify questune

Ran smtp.test61 at gmail.com
Wed Sep 9 14:26:02 UTC 2009

Hi ,
I didnt quite understand from the faqs what is better , gluster site say that
distributed is the new way to go while i see other people uses unify
Also i see that unify has many performance options while distributed
has very little options .
So the questune is what is the better way to go for a 10TB file system
with 5 servers
No replication made seens every server uses DRBD for better stablilty .

Another questune is seens distributed uses hash values
Is it safe to say that when creating a single server glusterfs
with the option to add more servers in the future
Is it better in terms of perfomance(seens its working ok)
to start with distributed 1 server mode and just add more in the future e.g

volume storage1
  type protocol/client
  option transport-type tcp
  option remote-host
  option remote-subvolume storage1

volume cluster
 type cluster/distribute
 option lookup-unhashed yes
 subvolumes storage1

volume writebehind
  type performance/write-behind
  option window-size 5MB
  subvolumes cluster

the valume cluster has only 1 subvolumes in distribute mode

Is that a good config for 1 server in terms of perormance ?

Ill arucitae also a goood recomendation to perfomance boost for FS
with many small files (mail etc..) seens my tests show a slower
results then nfs with dd of 1024 blocks while its getting much better
on 10240 .

Thanks ,

More information about the Gluster-users mailing list