You didn't provide your current config or what version you are running
but you may find this useful:
http://www.gluster.org/community/documentation/index.php/Gluster_3.2:_Tuning_Volume_Options
I'd be interested to hear what happens if you repeat your tests with
some performance settings changed. I will be configuring Gluster in an
HPC cluster with ~83 million files and I'm going to run into the 'lots
of small files on replica sucks' problem.
Also, here's my testing but it was on a very limited network (stuck on a
100Mb switch):
https://docs.google.com/document/d/1PpszoTd6CqQaaQKavSagw8_ySsEI_sXcCa5z4ubWgLU/edit
Jeff White - Linux/Unix Systems Engineer
University of Pittsburgh - CSSD
On 12/21/2011 03:09 AM, Ulli Horlacher wrote:> I have set up my first 2 node gluster with 1 GbE for testing.
> Thanks to very good documentation, it was easy.
>
> My aim is to use gluster for LXC VM replication.
> Write performance is a critical issue for me.
>
> Test case 1: copying 22000 files with 1 GB totally
>
> local disc/filesystem: 47 MB/s
> NFS3: 13 MB/s
> glusterfs: 4 MB/s
>
>
> Test case 2: writing one file with 1 GB
>
> local disc/filesystem: 123 MB/s
> NFS3: 95 MB/s
> glusterfs: 67 MB/s
>
>
> Especially the result of test case 1 is very disappointing. Is there a way
> of tuning it up?
>
>