[Gluster-users] Reference architecture

[Joe Julian]

On 01/31/2016 02:43 PM, Gandalf Corvotempesta wrote:
> Hi to all
> are there any reference architecture to use as base for a new environment ?
>
> We are trying to build a brand new object storage service for our customers.
> Actually we have 3 storage servers with 12 disks each. Each storage
> server will be used as Gluster node, one brick for each disks (2TB)
> with sharding.
>
> Are 10GbE really needed ? Actually we don't have any 10gbe switches
> but standard 1GB (i can create multiple LACP channels, if needed)
>
> Should I add 2 servers for gluster-swift as gateways? Which hardware ?
> Direct access to files (through glusterfs and not swift) will be made
> with FUSE and gluster driver across each client (mostly some virtual
> machines)
>
> There is no need to share files between gluster-switft and "plain"
> gluster but gluster cluster must be the same (same servers, same
> disks, same network.....). We don't want to build 2 different
> clusters.
>
> This new environment *MUST* be totally redundant with no SPOF.
>
> Any suggestions ?

 From my perspective, having helped people with their own implementation 
as well as going through many iterations of my own, there is no 
one-size-fits-all. They're all dependent on use case. Start with your 
needs and engineer something to exceed them.