SD-WAN with MPLS
Because SD-WAN is designed to utilize multiple independent transports from various service providers, integrating SD-WAN with an MPLS deployment is rather straightforward. There are numerous reasons why the enterprise may want an SD-WAN deployment while still utilizing an L2VPN or L3VPN MPLS deployment.
The first scenario is quite obvious: migration from MPLS to SD-WAN. In this scenario, the enterprise already maintains a WAN topology facilitated by their MPLS provider and plan to move to SD-WAN, perhaps to take advantage of less expensive broadband circuits. However, there are other scenarios where both the MPLS environment and the SD-WAN environment will coexist by design. For instance, the interconnection between data centers may be through an L2VPN offering that should be maintained even after the remaining WAN has migrated or deployed SD-WAN.
In most scenarios, the primary concern for design and implementation will be on proper route filtering. Depending on the routing protocols utilized within the environment, it is possible to inadvertently cause a routing loop through redistribution, as well as create suboptimal routing. For this reason, it is recommended that all best practices around route redistribution are strictly followed, including marking all prefixes that are redistributed from one protocol to another. This may be via OSPF and OMP tags, BGP communities, and so on.
SD-WAN has various built-in mechanisms to prevent route looping. For instance, when the OMP overlay AS has been configured, this ASN is added to the BGP as-path attribute when the SD-WAN Edge advertised the prefix into BGP. Additionally, when the SD-WAN Edge advertises an OMP prefix into OSPF, the down bit is set. The SD-WAN Edge works in a similar fashion to an MPLS PE node. However, without proper care, the mechanism used by SD-WAN to prevent route looping could be bypassed.