While the distributed model of today works fine, controller-based networks do have several advantages. For example, Ethernet loops in the network are not a problem. Loops in Ethernet generally pose a problem because there is no way to stop frames from going in one port and circling back around again. Spanning Tree deals with loops by shutting off one leg of the loop. The controller builds a model of the network and connections and sees multiple connections between devices as alternate paths that it can use to forward frames. Access controls can be applied at any point in the network where it makes sense. Quality of service can be achieved by load balancing paths through the network, ensuring that no choke points develop (or alerting admins when they do).
One of the basic design parameters is that OpenFlow can run on existing switch and router hardware and can replace the hardware forwarding mechanisms or run concurrently with them. More vendors are beginning to adopt OpenFlow. For example, NEC has announced an experimental switch, and other vendors are showing interest.
Considering that many IT shops aren't yet at a point at which they are considering multipath Ethernet such as TRILL, Shortest Path Bridging, Brocade's VCS or Cisco's FabricPath, OpenFlow could be a useful alternative for multipath Ethernet. As with any standard, any equipment vendor that supports OpenFlow can participate in the network, allowing you to have a multivendor network. If you are at Interop, check out the demo. It is sure to be interesting. In fact, it was advertised on Time Square. (It's not Shopped, if you're asking)