SD-WAN has yet to move off the drawing board at many enterprises, but it is never too early to start thinking beyond the initial deployment phase to a time when optimization becomes the order of the day.
And while the technology is purpose-built to incorporate all manner of intelligence, automation and other cutting-edge capabilities in the drive to continuously improve functionality, the question remains: how do you measure and manage the performance of a highly dynamic, virtualized wide area network?
One thing is certain, says Accedian’s Michael Rezek, todays collection of network monitoring solutions will not help the enterprise deliver on SD-WAN’s promises. The key problem is that solutions that focus on core MPLS or carrier infrastructure are designed to ensure Quality of Service (QoS), which might sound like a reasonable approach, but is not. QoS, after all, is based on metrics like performance degradation, connectivity, bandwidth availability and overall throughput. In an SD-WAN setting, however, problems can still exist between hardware and software or on the application layer, even if the network is functioning properly. In the control room, then, the administrator is seeing green lights, but the user is not getting adequate performance.
Some might say, so what?
Why should the network monitoring team be responsible for the entire application stack? Because in the virtual networking world, everything is interrelated. The physical layer must coordinate smoothly with the virtual layer, which must function with the operating layer, the management software layer, the application layer and even the data layer. To deliver a fully functioning SD-WAN, the enterprise needs to expand its focus beyond simple QoS to embrace Quality of Experience (QoE).
To be sure, says Infovista’s Ricardo Belmar, the enterprise need not give up on traditional QoS techniques in order to implement QoE. Things like forward error correction, path selection and traffic prioritization will always be necessary in order to optimize the network, and indeed, many of these functions will become automated. To realize QoE, however, you must accept the fact that one network environment is not optimal for all applications. So rather than simply deploying FEC the same way all the time, you’ll have to start optimizing delivery from the outset on a session-by-session basis, which can only be done once you have granular insight into the needs of each application. In this way, things like prioritization and routing can be based on business requirements, not simply data types or other arbitrary markers.
Part of this transition will involve making both QoS and QoE independent of the network. To accomplish this, several new capabilities must be introduced, such as implementing dynamic queues as optimized application overlays so that performance requirements can be mapped to key business objectives, and dynamically controlling queue sizes based on network conditions and traffic load.
Still, the complexity of emerging SD-WAN environments all but rules out the possibility of managing them manually. Instead, the network itself will have to become increasingly self-optimizing. Cato Networks took a step in this direction with its new Self-Healing SD-WAN platform, which gives the enterprises the tools to ensure high availability rather than relying on the carrier. Regardless of whether a failure occurs on the edge, the transport layer or anywhere else — or even if it is something routine like moving apps between clouds — the Cato SD-WAN will react and respond accordingly and automatically.
There are many ways to measure performance, of course. And when it comes to gauging customers’ satisfaction with their application experience, the metrics become even fuzzier. But ultimately, the proof will be in the pudding. That is, happy users will return to your service again and again, and unhappy ones, well, you can guess.
But no matter how the SD-WAN evolves over time, it is probably safe to say that the days when networking professionals could concern themselves solely with what’s happening on the network are coming to a close.
Arthur Cole is a freelance journalist with more than 25 years’ experience covering enterprise IT, telecommunications and other high-tech industries.