Do all servers have one base OS, like in RED HAT openstack architecture?

Solution 1:

OpenStack is a collection of software projects that create useful infrastructure out of raw compute resources. Typically you could create a few hundred small VMs out of what you have. Compare to your favorite public cloud: self service standard APIs, and the consumer does not have to concern themselves with hardware.

One giant OS instance running on many nodes is a bit different. That is a thing in the high performance computing area. Think TOP500 list. I don't think HPC is a common use case of OpenStack, HPC already has specialized tooling.

Red Hat has their own OpenStack distribution, packaging up these components and helping deploy them. (Help probably is required, I have heard the learning curve is steep.) Of course Red Hat has their own Linux distribution as a target platform to run everything on. RHEL at the bottom there is a logical, rather than physical relationship, as the "foundation" to build things on. There actually are many nodes where these services run. A dozen tiny RHEL logos would be difficult to read, so it just spans the diagram.

I am not qualified to say much about OpenStack Neutron networking. Neutron provides network services, in a broad, generic, and object focused way. Layer 2 and 3 connectivity, overlay encapsulation, firewall as a service, IPAM and DNS services, hardware offload. All components you may want in a modern multi tenant compute service. More fundamental than how to address a vNIC, how do you want to design your network?