SD-WAN is like an intelligent WAN accelerator that can centrally control WAN traffic in multiple network paths, including MPLS, VPN networks, and cellular data services LTE (see figure below). The SD-WAN controller makes multiple links look like a logical link, like the Ethernet link aggregation we have used for many years. Expensive links such as LTE are preserved with the decline of MPLS and VPN networks. The SD-WAN controller maximizes the efficiency of each path by using optimized WAN capabilities.
The SD-WAN controller determines the specific physical path of each packet based on user-defined policies, delay jitter, and packet loss in each path. The SD-WAN test path has two characteristics. The first is to add a wrapper header to the network traffic that passes normally in each link. The added header includes a sequence number and a timestamp, enabling the SD-WAN controller to determine the delay, delay jitter, and packet loss for the path. The second is to send a synthetic traffic containing the serial number and timestamp. Intelligent SD-WAN controllers will be combined in two ways to avoid reliance on normal network traffic.
The SD-WAN administrator configures different types of networking traffic and defines policies for each traffic class preferred path. For example, UC&C traffic will be identified by a small fixed UDP packet size, possibly using a generic UDP interface. The path selection strategy is designed to choose low latency jitter as the primary criterion, then low latency, and finally low packet loss rate. This strategy may result in packets using MPLS paths. Another strategy can be created for important business applications, choosing low latency and low packet loss rates as the primary criteria, and less important traffic will be downgraded and transmitted over another policy VPN network.