[Gluster-users] Dsitribution of load on servers

[Fred van Zwieten]

Laurent,

If you use the glusterfs fuse client for you're clients, all clients will
distribute across all servers, because the hash calculation will be done on
the client.

I you use Samba/Cifs or NFS, you normally have to connect to one server.
Here, you can spread the clients manually across the servers. If you want
to spread this automatically, you can set up CTDB, see
here<https://access.redhat.com/knowledge/docs/en-US/Red_Hat_Storage/2.0/html-single/Administration_Guide/index.html#chap-Administration_Guide-GlusterFS_Client>

Fred



On Sat, Dec 1, 2012 at 12:08 AM, Laurent Chouinard <
laurent.chouinard at ubisoft.com> wrote:

> Hi,
>
> I have searched the archive of this list and found some partial answers to
> my question, so I thought I'd try a direct question.
>
> In a nutshell, my goal is to achieve high transfer speed for multiple
> simultaneous clients using multiple servers. I am not trying to get
> incredible speed for one client using all servers.
>
> Assume a Gluster setup of 10 identical data nodes. I have a large set of
> data files that can be replicated n times (up to 10).
>
> At any point in time, a group of clients could request a set of those
> files, possibly the same set of files for everyone.
>
> Could Gluster manage that each client uses a different server so that, if
> everyone is getting the same file, then each client would use the maximum
> capacity of each server's hardware?
>
> In other words, using numbers:
> - 10 servers with 1 hard disk capable of 50MB/s
> - 10 clients requesting the same file, which is replicated 10 times
> - Speed for each client would be 50MB/s.
>
> Thanks,
>
> Laurent Chouinard
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20121202/48fb2eb1/attachment.html>