[Gluster-users] tons and tons of clients, oh my!

[Joe Landman]

On 10/20/2011 06:29 PM, Luis Cerezo wrote:
> Hello gluster-verse!
>
> I'm about to see if GlusterFS can handle a large amount of clients, this
> was not at all in plans when we initially selected and setup our current
> configuration.
>
> What sort of experience do you (the collective "you" as in y'all) have
> with a large client to storage brick server ratio? (~1330:1) Where do
> you see things going awnry?

We've had up to 300 clients per brick ratios in the past.  For large 
files, we saw lots of contention.  Contention drastically reduces 
performance.

> Most of this will be reads and locks on small files and dirs. our setup
> is 3xstorage node servers in a pure distribute config.

You do not want to even consider small files ... this is not a good use 
case for a cluster file system in general.  The only real way to deal 
with this is to put your backing file system on SSD or Flash, and have a 
very fast network.  Even then performance is not going to be that good.


-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
        http://scalableinformatics.com/sicluster
phone: +1 734 786 8423 x121
fax  : +1 866 888 3112
cell : +1 734 612 4615