If you are speaking about the Cinder backend, this is not something that Openstack will probably ever implement or accept.

The Cinder backend itself should be setup as redundant. Then when users request space on the block device they don't need to know about those things. They can be confident thier data is redundant.

Now you may have a user want to do something like attach two Cinder volumes from different availability zones and RAID them. That might work, I haven't tested that scenario. But at that point the RAID would be handled inside the instance itself.

What would a "RAID" feature in Openstack look like to you? Perhaps we can suggest a way to accomplish that.