[Gluster-users] 2 NAS design considerations

[Lars Hanke]

I'm planning to set up a distributed replicated Gluster installation on 
two NAS, where one NAS replicates the other. The NAS are not identical, 
not even in storage size. One has a 12 TB RAID5, the other is virgin and 
has 3x3TB disks. The entire data to store is currently less than 1TB, 
but I also want to keep time-machine like snapshots (not necessarily on 
gluster).

Concerning network load:

1) Will replication occur synchronously, i.e. will it essentially half 
the network bandwidth for writes, unless I use a second NIC for Gluster 
internal communication?

Since there are no constant writes, it would be a good idea to wait 
until the servers are idle.

2) Does it have any impact on performance whether I run a single gluster 
volume, or multiple volumes.

Concerning bricks:

I read several test setups where bricks were just directories in a 
larger volume. So it seems to work apart from the free space. Something 
like brick quota would be a nice feature, but I couldn't find any hint 
that this exists.

3) For a replicated system, is there any benefit of practical value to 
have a RAID on the node? The network is the limiting speed factor, and 
redundance should be given by gluster, right?

4) Are there any experiences with BTRFS?

Thanks for sharing your experiences,
  - lars.