Software-defined wide area network is one of the hottest items to hit WAN since MPLS. The technology offers many new features and, when coupled with the right usage scenario, has the potential to significantly reduce operational costs.
Most software-defined wide area network (SD-WAN) benefits focus around the ability to replace or augment dedicated WAN circuits with Internet-based broadband circuits. With the move away from traditional dedicated circuits, a myriad of new features and functions can make the WAN more robust. Tie this together with simpler deployment models and easier management, and SD-WAN starts to sound more appealing.
Prior to SD-WAN architecture, there were a couple of options for site connectivity. Traditional WAN circuits provide dedicated private and predictable connectivity on WAN, but circuits of this type are typically expensive, and their deployment and modification can take weeks to months. A more inexpensive option is to deploy site-to-site virtual private networks across existing internet circuits. While this is feasible with a small number of sites, it could become complicated to manage a full mesh of VPN-connected sites.
SD-WAN architecture aims to solve many of the problems with previous iterations of WAN technology through increased flexibility. Since SD-WAN technology is based on an overlay, it can be provisioned over any type of WAN connectivity: dedicated or Internet-based circuits. In addition, SD-WAN benefits include provisioning and management that is abstracted into a controller and configured from a central location. Even if you're comfortable with the existing Multiprotocol Label Switching (MPLS) network, applications around segmentation and encryption make using SD-WAN over MPLS more attractive.
What makes SD-WAN architecture a serious competitor in this space is its features, particularly those related to path aggregation or bonding, path selection and path conditioning. In traditional WAN, dynamic routing protocols move traffic on and off of links depending on their current state. While this works well in most scenarios, it doesn't provide much flexibility. SD-WAN abstracts any existing circuits, or underlay networks, into a single logical WAN connection. We can then classify traffic by connection and even aggregate links of disparate types together. The SD-WAN device manages the underlay and allows the network team to focus on delivering connectivity. Some SD-WAN vendors offer more advanced features such as guaranteed packet delivery, packet-based load balancing and built-in WAN optimization.
Coupling SD-WAN with a WAN optimization technology is a further example of SD-WAN's benefits. Sites that traditionally had a rack full of gear to provide connectivity can now likely be serviced with a couple of small SD-WAN appliances and a few wireless access points. And while a rip-and-replace model may not make sense from a cost perspective, site refreshes and new builds may be good places to take advantage of this technology. Building a site can be as easy as ordering an Internet circuit and shipping a small SD-WAN appliance. Not only is the deployment faster, but managing and troubleshooting the WAN is greatly simplified by having a central controller for all of the endpoints, since changes are quickly and easily pushed out to the entire WAN. The controller also enables troubleshooting from a network perspective, offering aggregated metrics and historical trending data on key metrics.
One of SD-WAN's benefits is its consumable WAN virtualization, or its ability to easily segment users as they traverse WAN. Previously, this wasn't easily done without layering a technology such as dynamic multipoint VPN on top of an existing WAN. For consumers that are looking to extend isolated networks from the data center out to WAN connected sites, SD-WAN can potentially help solve the WAN piece of that puzzle.
Complications from SD-WAN
While SD-WAN aims to solve many problems, it does introduce some of its own. Users looking to solely use Internet-based connectivity are at the mercy of the Internet, although there are features of SD-WAN that can make internet circuits behave much like dedicated WAN circuits.
Most SD-WAN appliances only accept Ethernet-based connections, ruling out any deployment across legacy time-division multiplexing-based circuits without some kind of translation to Ethernet.
While there are some limitations of SD-WAN, it is here to stay. Its flexibility and ease of management make SD-WAN applicable in almost any kind of WAN deployment, whether in an enterprise data center infrastructure, colocation provider, cloud service provider or other user.
What's driving virtual WANs?
A look at the vWAN vendor landscape
When WAN optimization goes bad