Most people thinking about SD-WAN for voice-over-IP (VoIP) fall into one of two camps:
- The optimists: VoIP will work great over my SD-WAN—maybe even better than what I have now. Where do I sign up?
- The pessimists: Voice services are just too sensitive and too important to my business to use anything other than dedicated MPLS links.
Who’s closer to the truth? Both sides actually miss the mark. SD-WAN solutions can handle VoIP traffic. But just shifting to SD-WAN won’t automatically make voice quality at branches better—and could even make it worse. The question here isn’t whether or not to use SD-WAN. When it comes to SD-WAN for VoIP, the solution you choose makes all the difference.
MPLS or DIA?
Historically, most businesses running VoIP used dedicated MPLS links at branches, because voice is highly sensitive to things like delay and jitter. MPLS supports end-to-end traffic prioritization through class-of-service (CoS) and quality-of-service (QoS) techniques, where direct internet access (DIA) links can’t.
But guess what: you don’t actually have to choose between MPLS or DIA. Many businesses adopting SD-WAN augment rather than replace their MPLS connections. Theoretically, this is great for voice applications, because you now have more capacity to add more users and VoIP calls. First though, you have to solve the quality question. And for that, you’ll need to look deep inside your SD-WAN.
Evaluating SD-WAN Solutions for VoIP
First-generation SD-WAN solutions route application traffic over the best available connection, based on application profiles. That’s fine as far as it goes, but if your solution automatically sends all voice traffic over MPLS instead of internet (and many work that way), you’re undermining the benefits of your SD-WAN investment. You can maintain consistent quality, but if you bought SD-WAN to add more users and calls, well, you have the same VoIP capacity as before.
To use some of your DIA capacity for VoIP, you have to be able to control voice quality end-to-end, and first-generation SD-WAN solutions can’t—even when the marketing says otherwise. Most solutions (and practically all first-generation solutions) can only apply prioritization as traffic enters or leaves the branch. The entire time traffic traverses the internet between sites, you have no control at all. To guarantee voice quality, you need an SD-WAN that adds its own layer of end-to-end QoS across the various network links it uses, and apply different rules for different types of traffic.
For our Ipanema SD-WAN solution, Infovista looked closely at leading voice and collaboration vendors to understand these applications and their traffic profiles. We worked closely with Microsoft to build a deeper level of traffic handling intelligence within Skype for Business. A Skype for Business session, for example, can carry voice traffic, video traffic, data-sharing traffic or a mix of all three—and that mix varies from session to session, user to user. Ipanema applies different QoS rules depending on traffic type to optimize quality. It can even send different parts of a session over different links and reassemble them at the other end of the session.
To get that kind of quality control, you need more than just application profiles. You need deep application visibility and control on a session-by-session basis, and the ability to dynamically adjust routing decisions and QoS queues on the fly based on the real-time capacity and quality of all available links. Then you need to do this for every new call and user! It’s a much more sophisticated intelligence than first-generation SD-WAN solutions (and even many current-generation solutions) provide. But if you’re going to guarantee great VoIP quality over SD-WAN, including direct Internet access, it’s essential.