Basically, all stored data in ceph is stored as rados objects, independent of the application you're using (CephFS, RGW, RBD). For OpenStack you typically use RBD (Rados Block Device) which appear as block devices to the clients (instances), but they are still stored as many objects within ceph. That's how ceph scales up: many different objects of the same rbd can be queried on different OSDs (Object Storage Daemon). There's no such thing as an option what to choose from ceph, it's completely up to you how you use it. You only have to create the specific pools and users (clients) and tell your applications how to connect to ceph, which pool to use and which client it is. Of course, there's a lot more to be aware of (redundancy, failover scenarios, crush rules etc.) but maybe this is enough to get a grasp of it. I'd recommend to give it a try with ceph, there are several ways to install minimal clusters without much effort.

OpenStack exists, Ceph exists, but there is no such thing as OpenStack Ceph.

I think your questions are answered by this web page: https://docs.ceph.com/docs/master/arc.... In short, Ceph is Object Storage at its core, but also presents block devices through its RBD interface and a distributed filesystem through CephFS. Both block devices and filesystems are implemented using objects.

However, I don't think files are stored as objects. Filesystems are.

I am not sure if you can install Ceph without RBD and CephFS. Why would you? To learn more about installing Ceph: https://docs.ceph.com/docs/master/ins....