OVS is simply not a router. It lacks L3 interfaces and IP protocols.
You'll need a VM instance (or logical interfaces on the local Linux stack) to act as your vRouter. There are many free/open-source and commercial options for this function, but perhaps the easiest (yet feature-poor) option is to eliminate OVS completely and just use native Linux Bridging, which allows you to add IP numbering to your bridge interfaces.
Here is a VERY simple example in Ubuntu Server 14.04 LTS:
auto br0
iface br0 inet static
bridge_ports em1
bridge_stp off
bridge_fd 0
bridge_maxwait 0
address 192.168.1.200
netmask 255.255.255.0
gateway 192.168.1.1
dns-nameservers 192.168.1.1
up /sbin/ifconfig $IFACE up || /bin/true
You can add multiple physical interfaces to this same bridge, but based on your description you likely want two separate bridge interfaces, each associated with a different physical interface, each with its own IP address.
Routing is a larger conversation. Note that you'll only want ONE bridge with a default route, normally. Hosts/VMs on either side will need to know the route to the other side/subnet via static routes, or perhaps their def.gwy host will handle that static route for them. Again, this is a larger/different conversation.
Don't know what you want indeed. If you want two os-clusters communicate with each, I assume, you can simple use floating ip and configura three-layer routing policy for the two clusters. There is no need to use ovs and linux network can do this well.
This is an experiment for the role of SDN in wm migration between two layer3 domains. This is why I need to do that in a way I explained. I know there are many easier way to do that but I have to do it as I explained before