Hello All,

I am about deploying a new Ovirt system for our developers that we plan to be hyperconverged environment Node based.

The system workload would be mostly used for builders that compiling our code , which involves with lots of small files and intensive IO.

I plan to build two glustered volume "layers" one based on sas drives for OS spin on, and second for Nvme based for intensive IO.

I would expect that the system will be resilient/high availability and in the same time give enough  good IO request for vm builders that will be at least 6 to 8 vm guests.

The system hardware would be as follows:

chassis: 4x HP DL380 gen8

cpu: 2x e5-2690v2

memory:256GB

Disks:12x 1.2TB sas 10k disks , 2 mirror for os (or using usk kingstone 2x 128gb mirror) rest for vm os volume.

Nvme: 2x960GB Kingstone KC1000  for builders compiling source code

Network: 4 ports  Intel 10Gbit/s SFP +

Given above configuration and theory,  my question would be what would be best practice in terms of Gluster configuration Distributed,Replicated,Distributed Replicated,Dispersed,Distributed Dispersed?

What is the suggestion for hardware raid type 5 or 6 , or use ZFS?

Network nodes communication , i intend to use 3 ports for storage communication and one port for guests network , my question regarding Gluster inter communication , what is better would i gail from 3x 10G LACP or 1 network for each gluster volume?

Please advice

Thanks

--