A Self-Driving, Self-Healing SD-WAN: Achieving Intent-Based Networking In The Enterprise

The enterprise application landscape is spreading out. Not only have users been distributed globally for years, but now, the applications are moving out of the private enterprise data center. Organizations are shifting from on-premise application server environments to a multi-cloud world, leveraging SaaS and cloud services from multiple providers and deploying more real-time applications like Unified Communications. With this shift, IT teams now have far less control over the wide-area network (WAN) connecting users to these applications than they did with their private MPLS WANs.

The old model of a fixed transportation infrastructure funneling all traffic to a single data center is dead. Put another way, rather than trains confined to tracks to move our “stuff” around, we need cars. While an older manual transmission car can go wherever needed with a savvy driver (or network administrator) at the wheel shifting gears and constantly monitoring vehicle health, and an automatic transmission would certainly make things easier on the driver, a self-driving car will ‘take the wheel,’ find and follow the best route, avoiding traffic jams and anticipating issues before the check engine light comes on.

Standard Software-Defined Wide Area Networking (SD-WAN) offerings give enterprises the ability to forgo the IT labor-intensive processes of shifting gears à la a manual transmission in favor of a system that offers the possibility of branch office consolidation and is easier to manage centrally, with capabilities like zero-touch provisioning. But while that move from a manual transmission to an automatic transmission is important, it only addresses 20%-25% of the problem, especially if part of the goal is to safely and reliably support applications migrating to the cloud.

By contrast, a self-driving, self-healing SD-WAN is capable of delivering on intent-based networking, where network administrators can set high-level policies and have the SD-WAN technology deliver application traffic reliably and with high Quality of Experience (QoE) to its destination. Beyond streamlining network operations, this enables IT teams to reallocate resources to other pressing challenges, and, it means mission-critical business applications and processes can be run in the cloud and accessed over Internet connections without sacrificing the network reliability, and QoE enterprises have come to expect from their MPLS WANs.

SD-WAN, Take the Wheel

Most current SD-WAN solutions force network administrators to map application flows to specific preferred networks, and then (futilely) guess the packet loss or jitter thresholds to use to “session-steer” traffic to an alternate network. In contrast, a self-driving, self-healing  SD-WAN allows network administrators to set high-level application policies, and then the SD-WAN technology takes the wheel, applying intelligence to make per-packet forwarding decisions and react sub-second to problems detected – delivering traffic to its destination reliably and with high QoE based on the state of all the network connections right now, much like advanced mapping applications reroute cars to avoid congestion on the road. 

Map a Route for Success

For organizations seeking to make Intent-based Networking a reality today, there are specific capabilities necessary to deliver a self-driving, self-healing SD-WAN:

One-Way Only, All the Time: Continuous unidirectional measurement of all traffic between locations is the key. Continuous traffic measurement of all traffic, rather than just heartbeats or round-trip pings, radically improves the accuracy and timeliness of detecting WAN problems. Unidirectional measurement is important because traffic jams usually occur in one direction, not the other. And of course, for real-time traffic like VoIP and videoconferencing, the unidirectional performance is what’s relevant.

All Roads Lead to Rome – So Take Advantage of Them: The Internet is a network of networks. With multiple Internet connections, there is more than one network path possible for each WAN link. MPLS usually offers multiple paths for the multiple Classes of Service (CoS) purchased. A self-driving, self-healing SD-WAN takes full advantage of all these paths to deliver optimal reliability and QoE, measuring all possible paths between any pair of locations. This approach carefully considers business policy intent, putting the most important and delay-sensitive traffic on the currently best-performing paths, getting important traffic to the other end of the WAN quickly and reliably, and even using all WAN bandwidth between locations for a single flow.

Don’t Step Across the Threshold: The fractal nature of congestion-based loss makes setting meaningful thresholds to determine when to move traffic off of a connection impossible. SD-WAN solutions that “allow” administrators the “privilege” of setting thresholds to determine when to steer traffic is comparable to placing a 1950s crank engine into a modern car. Congestion-based loss on shared IP networks is unpredictable in duration, and so thresholding is a poor way to detect when a WAN thruway is no longer reliable. And this fatal weakness is further amplified if the SD-WAN lacks a continuous unidirectional measurement scheme.

Monitor for Closures (and Potholes): To keep traffic on course, network administrators should demand solutions offering “fast-twitch” proprietary network path heuristics. This close cousin of self-discovery is a co-pilot in the self-driving WAN. Heuristics intelligently watch when a ‘good’ WAN road has gone ‘dead,’ or for less obvious signs, it has gone ‘bad’ (a much harder task to determine). Enterprises should ensure ‘bad’ WAN connections are quickly avoided until the danger is gone unless no other ‘good’ roads are available.

With the right self-driving, self-healing SD-WAN solution in place, enterprises can make intent-based networking a reality today. Empowering IT administrators to deliver high QoE while setting high-level policies, rather than mapping applications to networks and fiddling with impossible-to-set thresholds, is the way to deliver better quality user experiences across countless applications. After all, no one likes sitting in a traffic jam on the road, so why risk that with your WAN, especially as your applications migrate to the cloud?