Enterprise SD WAN Architecture and Implementation
SD-WAN Architecture Fundamentals
Breaking Up the Router
Traditional wide-area networks (WANs) have a straightforward, but rigid, design. Each router or switch is an all-in-one device. It decides where traffic should go (the control plane), it physically forwards that traffic (the data plane), and you configure it by logging into its individual command line (the management plane). This works, but it doesn't scale well. Making a network-wide policy change means logging into hundreds or thousands of devices one by one, a recipe for errors and long nights.
Software-Defined WAN, or SD-WAN, flips this model on its head. Instead of keeping all the intelligence on each individual device, it separates these core functions into distinct, logical layers. The core principle is decoupling the network's brain from its body.
The core principle of SDN is to decouple the network's brain (the control plane) from its body (the data plane).
This separation allows for centralized management and automation that's impossible with traditional hardware. You can push a new security policy or application routing rule from a single dashboard, and it's instantly applied across the entire network. This architectural shift is what gives SD-WAN its agility and power.
The Underlay and the Overlay
To understand how SD-WAN operates, you need to think in two layers: the underlay and the overlay.
The underlay network is the physical infrastructure. It's the collection of MPLS circuits, broadband internet connections, 4G/5G links, and any other transport medium that connects your sites. The underlay's only job is to provide basic IP connectivity. It moves packets from point A to point B, but it has no special intelligence about the traffic it's carrying.
The overlay network is the virtual, intelligent fabric that SD-WAN creates on top of the underlay. This is where the magic happens. The overlay consists of secure, encrypted tunnels (often using IPsec) that are established between all the SD-WAN edge devices. It's within this overlay that all the advanced routing decisions, security policies, and application-aware traffic steering take place. The overlay doesn't care what physical links are in the underlay; it sees them all as just available paths to send traffic over, choosing the best one based on real-time performance.
The Four Planes
The decoupled architecture of SD-WAN is best understood through its four distinct planes of operation. We'll use the popular Cisco Viptela solution as a concrete example, where each plane has a corresponding component.
1. Orchestration Plane (vBond Orchestrator) The orchestrator is the first point of contact for any new edge device joining the fabric. Think of it as the network's bouncer. Its primary job is to authenticate devices, ensuring only legitimate hardware can join. It checks a device's serial number against a pre-approved and, if it matches, tells the device where to find the network controller (vSmart). This process enables (ZTP), where a new device can be shipped to a remote office, plugged in, and it will automatically configure itself and join the network without any manual setup.
2. Management Plane (vManage) This is your single pane of glass for the entire WAN. The management plane provides the graphical user interface (GUI) and APIs for configuring, monitoring, and troubleshooting the network. Instead of logging into individual routers, you log into one centralized system, vManage. Here, you create policies for routing, security, and quality of service using templates. These templates are then pushed out to the controllers and edge devices, ensuring consistent configuration everywhere.
3. Control Plane (vSmart Controller) The controller is the centralized brain of the SD-WAN fabric. While vManage is for human administrators, vSmart is for the devices themselves. It receives operational information from the edge devices, like link quality and path availability. Based on the policies configured in vManage, the calculates the best paths for application traffic and distributes these routing policies and security information to all the edge devices. It's responsible for maintaining the topology of the overlay network.
4. Data Plane (vEdge/cEdge Routers) Finally, the data plane consists of the edge routers at your branch offices, data centers, and cloud environments. These are the workhorses that actually forward packets. They receive their forwarding instructions and policies from the vSmart controller and execute them. They constantly monitor the performance of the underlay paths (measuring latency, jitter, and packet loss) and report this information back to the control plane. This is the only plane that is physically distributed across all your sites.
Let's review these concepts before moving on.
Now, test your knowledge of how these planes interact.
What is the core architectural principle that distinguishes SD-WAN from traditional WANs?
In an SD-WAN architecture, the intelligent, virtual network built on top of the physical connections is known as the ___________ network.
By separating these functions, SD-WAN provides a level of centralized control and automation that is simply not possible with legacy network architectures.