Will Year 2021 Put SD-WAN in Autopilot?

February 11, 2021

By Special Guest
Ashish Jain, CEO & Co-Founder, KAIROS Pulse,

SD-WAN solutions and AI are poised to bring significant benefits to MSPs and businesses in 2021 as they continue addressing the remote working challenges brought on by COVID-19. SD-WAN brought monumental success in 2020 by powering the transitions for countless remote office workers, as SD-WAN vendors and MSPs alike rose to the challenge to manually support this unexpected high growth ramp. Success was hard-fought, but the operations teams found ways to keep home-workers connected. Now, network providers are looking for ways to improve the operations of their networks. With more than 160,000 SD-WAN sites operational at the end of 2019 (before the pandemic hit the network), MSPs are looking for ways to efficiently manage the sudden flood of manual data analysis tasks so that their networks operate smoothly, and their customers remain satisfied.

The ability for SD-WAN solutions to use multiple links as a single connection helps MSPs keep enterprise data flowing to their cloud services when some connections are underperforming. By incorporating AI in an SD-WAN solution, the network operations team can get a more accurate handle regarding the performance of the network and the applications.

Predictive Analysis in Today’s Solutions
With the metrics and KPIs in today’s processing, operations teams rely on manual expertise and tracking to determine how a network is operating. For example, the gradual trend of a growing usage of an application may not be directly apparent to staff, resulting in periodic impacts that results in increasing trouble tickets due to poor performance, but go away right as the team begins troubleshooting.

Today’s human-based “tracking” in these large networks is not optimal. But with AI-powered metrics and KPIs, support teams gain instant insights from AI-based alerts that simply cannot be deduced in real-time from “manually” processing. AI-infused SD-WAN with application-level monitoring can automatically recognize operational trends by analyzing current operations with similar situations. These AI-powered analyses can perceive even the smallest of anomalies that may are not even currently impact a site and alert the operational teams of potential issues. These warnings from automated processing help the operations team proactively adjust the network traffic flows and activity to avoid reduced performance.

Putting SD-WAN on Autopilot
The next step, which may occur as soon as 2021, is to have the results from the analysis drive automated changes into the cloud-controlled SD-WAN. The automated identification of issues coupled with the ability the provision changes on the fly is creating an optimal solution. This intelligent system can create a “self-flying” network. But this pilot needs to be able to monitor performance, manage capacity, optimize traffic flows, identify and resolve security concerns, and capture, analyze, and troubleshoot network anomalies with minimal human action. Much like today’s autopilot functions on state-of-the-art airplanes, the SD-WAN autopilot needs to interact between the systems of the network.

To achieve this level of automation, the SD-WAN must have:

  1. Cloud-accessed metrics of links across the SD-WAN – with access to performance data across all sites, traffic flows, and links.
  2. AI-powered analysis of historical data – AI-analysis and machine learning algorithms allow the system to identify the available capacity in real-time that meets latency, jitter, and packet-loss metrics for meeting SLAs.
  3. Automated orchestration across the SD-WAN – the automated management and allocation of connectivity resources will surely be able to take the network beyond 99.999% availability.

Getting Ready for Takeoff
AI and Machine-Learning are being used in many services today, so the expertise needed for creating the foundation for building these autonomous networks is well under construction. While today’s systems are not yet linking the power of AI-powered analysis with the cloud-orchestration mechanisms, this work is surely underway. Service providers and enterprises will soon have the automated network systems that keep their services operating at top performance. However, they will still need the support of MSPs to perform the manual tasks for plugging devices into the network, much like pilots still have to control the airplane before takeoff and after landing and when conditions require them to change course.