[Gluster-users] New architecture: some advice needed

[Xavier Hernandez]

For detailed information, please look at this email because it contains 
much information about things to be aware of:

http://www.gluster.org/pipermail/gluster-users/2015-January/020059.html

As a general rule, having few big replicas will reduce the time needed 
to rebalance the volume when it's expanded with more bricks, but the 
time needed to heal one brick will be higher.

On the other hand, many small replicas will be healed faster, but 
rebalance will take more time.

Regarding RDMA, it's considered tech preview now. In theory it will be 
fully supported starting with 3.6.2.

Xavi

On 01/09/2015 09:35 PM, Gandalf Corvotempesta wrote:
> 2014-12-25 21:57 GMT+01:00 Gandalf Corvotempesta
> <gandalf.corvotempesta at gmail.com>:
>> And what about hardware? Do you have suggestions?
>> Actually I have 3 of these supermicro server: 6027R-E1R12T
>> All with 2 SSD and 1 Infiniband DDR dualport.
>
> Bump...
>