We are all learning the acronym SD-WAN, but to layer 2 and 3 OSI model aficionados, it can be hard to accept how an upstart layer 7 software application can potentially replace a stalwart layer 3 technology like multiprotocol label switching (MPLS). Which begs the question: Is SD-WAN a contender, or just a pretender in the future generation of networks?
MPLS: The Incumbent
When MPLS was introduced, medium and large businesses rapidly adopted the technology. Deploying an MPLS network to connect offices together provides users with the ability to have a predictable, secure, and high preforming environment. This is done by bringing all your data through a single provider and allowing them to prioritize your traffic at layer 2 or layer 3 using QoS.
Providers can provide QoS through specifically designed networks with logical separation using labels to identify traffic, VRFs, and VRF tags to separate the customer traffic. They then allow customers to apply QoS tags to traffic types to reach higher levels of availability.
Let’s look at how this works in more detail, as this is the meat of what customers care about.
When designing MPLS, it is important to determine how you want to provide this QoS. The first thing to do is identify traffic of similar kind and then a degree of excellence. In a typical deployment, 4 quality queues are seen as defined below:
- Real-time traffic – This is typically voice, video, and/or VDI.
- Business Critical – This includes applications required for a company to do business. Examples might be an EMR for a medical company or credit card transactions for a retail company.
- Business Important – This might be an intranet, active directory, email, or other application that is not sensitive to jitter, latency, or retransmits.
- General Traffic – This queue includes all other traffic.
Once customer traffic has been identified and tagged, the degree of excellence of those tagging/groupings must be defined. Most carriers offer a different service level for each queue and different bandwidth guaranteed to each traffic type across the entire MPLS network. As companies continued to converge, however, holes in this design were found. This brought forth the creation of SD-WAN.
SD-WAN: The Contender
Instead of using QoS (layer 2 or layer 3), SD-WAN uses software definitions. With MPLS, you must either use only a source or destination IP address or port. That would be the same as saying you can only relieve traffic congestion from where you started, where you are going, or based on the type of car you have. This would not be very helpful in a major situation where most traffic is destined for the same general area. But what if you could also consider important factors, like ensuring doctors and first responders were never caught in traffic? That one additional factor could change the landscape in most cities and potentially save lives. In terms of the company network, that would be the same as having a real-time queue.
This is what SD-WAN does to QoS. Video traffic can be prioritized to YouTube over Netflix, for example, or Office 365 email over personal Gmail accounts.
Not Beholden to a Single Carrier
With MPLS, the entire environment must be controlled by a single provider. This means that you cannot choose the most cost effective solutions for each location, which doesn’t work for customers with large geographical environments. SD-WAN allows the most cost effective solution per location be chosen.
Significant Resilience Improvements
With MPLS, providing redundancy is very difficult as MPLS routing uses private IP. For the secondary connection, most customers choose to use a VPN over an internet connection. They then need a device that is intelligent enough to build the VPN and handle routing between the two solutions. Even then, routing with just layer 3 protocols doesn’t allow different levels of resilience or performance increases. Typically, customers settle on having the VPN be active/inactive (or hot/cold), meaning that the other connection is ONLY used when there is a total failure of the MPLS.
With SD-WAN, the world of high-performing resilient networks comes with a lot of options. You can define and build it so that both connections are used, called active/active or hot/hot. You can prioritize based on application type; for example, real-time traffic can be forced down a high-quality connection with SLAs associated with it, and general web traffic can be sent down a more cost effective connection. Other business critical or important traffic can be bundled to go out both connections, maximizing total bandwidth.
The flexibility and design options increase exponentially with the introduction of SD-WAN. MPLS still has its place and will for years to come, but for many customers SD-WAN provides a great alternative that may help them meet their internal goals.