When we first presented at SD-WAN Summit in 2019 about the applicability of AI/ML on SD-WAN networks, I have to admit that the essence of our pitch was coming from a strong intuition rather than practical experience in that space. This is because most of our production deployments at that time were in WAN and Datacenter segments.
However, when I look back at those old slides today when the new edition SD-WAN and SASE 2021 is starting, I am really surprised by how well they resonate with our last two years of deployment in enterprise and MSP SD-WAN infrastructures.
They are at least three operational use cases that have been confirmed by production deployments using AI/ML in SD-WAN SASE context:
- SD-WAN infrastructure grey failure detection: the objective is to be able to detect among noisy events abnormal patterns highlighting SD-WAN control plane or data plane instabilities before a failure occurs
- Overlay/underlay correlation: MSPs need to isolate SD-WAN misbehaviors from underlay connectivity grey failures. Enterprises are even more demanding as they want to automatically determine faulty third-party ISP connectivity service without having any access or monitoring data provided by these underlay networks
- Policy tuning helper: application policies and associated SLA thresholds are challenging to configure and are often not optimized to take best advantage of network conditions. There is a need for automated assistance with tuning insights.
For each of these cases, we have learned during the last two years where ML is the most actionable , and sometimes the details that we have discovered have surprised us.
We will go through these use cases in the coming weeks, as we are today at a stage where AI with machine learning is starting to play a pivotal role in the SD-WAN paradigm, and this is just the beginning.
Please keep in touch.
