What's the difference between "underlay" and "overlay" links in SD-WAN?

A quick explanation for underlay and overlay networks.

What's the difference between "underlay" and "overlay" links in SD-WAN?
Photo by Jordan Harrison / Unsplash
💡
For the purpose of this article/topic, I'm simplifying some of the concepts. Certain vendors have proprietary innovations and particular mindsets. Take this post with a small grain of salt.

In SD-WAN, the terms "underlay" and "overlay" are absolutely everywhere. What do those terms actually mean though? Let's unpack it.

An underlay refers to the physical links provided by an ISP. These days, there are a lot of choices but some examples could include:

  • Cable broadband internet
  • DSL (yes, that still exists 😄)
  • Fiber
  • MPLS
  • Wireless broadband (4G, 5G, etc)
  • Etc, etc, etc.

There are other examples of course, but hopefully this gets the point across. Think of underlays as part of the ISP physical infrastructure. Because that infrastructure is owned by the ISP, it's fairly rigid and there are practical limitations restricting how much consumers/administrators can modify the underlay itself.

An overlay refers to virtual links that are built on top of underlays. Overlays use the underlays as pathways to traverse and get from one place to another. Examples of overlays can include:

  • IPsec
  • GRE
  • IP-in-IP

The overlay traffic is often encapsulated and that means that it has more freedom and flexibility in how the data leave; it's no longer restricted to the confines of one WAN.

Traffic example

An example might help. Think of it this way: imagine commuting from home to the office. An underlay is like a physical road/highway – it's a defined piece of infrastructure that's fairly rigid. You as a bicyclist/driver can't really change the physical attributes of the road. The road is always going to be the road.

On the other hand, an overlay is like a GPS navigation route that guides you from home to work. The GPS route isn't required to use any one specific path; it can choose different paths for different scenarios/needs/conditions. If I'm bicyclist, maybe it's not a smart idea for me to be riding my bike on the interstate. Conversely, if I'm driving an 18-wheeler semi-truck, maybe the interstate makes more sense. What about timing? If it's 8:00 AM during peak traffic conditions, maybe I need to use different roads... etc.

Sorta make sense? Cool. Here's some additional reading material on the topic:

Understanding Underlay and Overlay networks
Hello community, First of all if that question has already been asked i’m apologizing in advance for i have not found it in the forum. I’m beginning a new step of my career in networks and that takes me to DC architectures design. I am trying to catch up with all the notions that i lack (and tha…
Underlay vs Overlay Routing
The overlay fabric solves most of these inefficiencies by changing the traditional routing concept of a next-hop IP address with a next-hop TLOC. As vEdges build overlay tunnels between their WAN tunnel endpoints (TLOCs), they advertise the site-local networks as reachable via their local TLOCs.
SD-WAN Overlay versus Underlay Networks: Key Differences Explained | NSC