Let us assume that BGP is used between all the following scaled down nodes. The spine-1 is used as an aggregation router for leaf-1 and leaf-2. Each of the leaf router is top of rack switch that advertises a /24; in this case, leaf-2 is advertising 10.20.1.0/24. We would like to migrate leaf-2 from traditional IOS-based device to OpenFlow device with Ryu controller.

Of course, your network might look different than this. This topology was picked because:

• It is an easy segmentation in your mind. You can easily point to the leaf-2 rack and think of it as an OpenFlow rack.
• It is a clean swap and is easy to roll back. In other words, you are replacing a physical hardware for the same function. In the event of a rollback, at worst, you can always just take out the new router and put the old one back in.
• The logical addressing is clean. You can be on high alert for any server reporting error in the 10.20.1.0/24 block upon migration.

Regardless of your scenario, these three points would be my recommendation for picking a spot in your network as a starting point for migration. In the next step, let's see how we can mimic this scenario with Mininet, VIRL, and our virtual machine.