[Gluster-devel] is it a bug? (unify+rr)

Yi Ling lingyi.pro at gmail.com
Sat Aug 1 02:48:53 UTC 2009


2009/8/1 Amar Tumballi <amar at gluster.com>

> 'cluster/distribute' is just a symlink to 'cluster/dht', which works on
> different design internally compared to 'cluster/unify' which needs a
> centralized 'namespace' volume to work. We have noticed that the 'scaling'
> is not possible with 'cluster/unify' and have decided to not continue any
> more development on cluster/unify design, instead any more enhancement of
> glusterfs will happen only on 'cluster/distribute' translator.
> -Amar
>
> there are two qeustions:
(1) what's the functionality of "cluster/unify" in current release? in my
opinion, "cluster/unify" is a translator for scaling, and now it doesn't
function like that. is it right?

(2) whether will "gluster/unify" still exist in the future release? and
could i suppose that the centralized namespace needed by "cluster/unify" is
something that functions like metadata in current release?

on page 5 of the technical presentation (
http://ftp.zresearch.com/pub/gluster/glusterfs/talks/Z/GlusterFS.pdf) ,there
is a "unify" block on client side. so im gussing that "unify" is a essential
component of client, with its namespace volume, shared by all clients,
instead of metadata which traditional cluster filesystems keep, like Lustre.

do you agree with me ?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20090801/a6d2fca4/attachment-0003.html>


More information about the Gluster-devel mailing list