Written by Dan Pitt
Dan Pitt is Executive Director of the Open Networking Foundation, joining on its public launch in March 2011.
Overlay and underlay designs have different unique consequences on the network topography. An overlay network places a selected flow based path at the end points in its software implementations, then it essentially tunnels through the network. On the other hand, underlay networks put a software defined network at every point in your network and it lowers the cost of those intermediate components, which allows you to software to define their operation.
An overlay network puts sort of your flow based path selection at the end points and often in hypervisors, in its software implementations then it basically tunnels through the network. That's maybe your only choice if you don't control what's in the rest of the network.
If you do control the network, you're going to want to have control over how your flows are treated every point in the network, because you have priorities. You have security concerns. You have compliance requirements, might even have legislation that mandate how you treat your traffic.
So, you want to be able to do it throughout your network. You want to be able to cache your videos in certain places. You want to be able to restrict students from accessing parts of the network during certain parts of the day if you're running a campus network.
Consequently, you will want to manage all points in the network and that's what the underlay is. It puts a real software defined network at every point in your network and it lowers the cost of those intermediate components and allows you to software to define their operation. It just makes them high performance packet processors.