AIOps has been making the rounds in networking marketing departments for a few years now. The big promise has been that AI is going to perform analytical thinking for us and, when things are going wrong, make life easier for engineers. The reality has been somewhat different. We’ve gotten lots of statistical analysis tools branded as AIOps–not really AI, but simpler math models that can detect when certain metrics deviate too far from normal. Fancy threshold alerting was mostly all that was behind the curtain. No machine learning. No artificial intelligence. I am very happy to report that over the last year, we’ve seen AIOps applied to networking in a meaningful way. Tools have come to market that use machine learning and artificial intelligence to analyze vast amounts of telemetry, detect patterns in the noise, surface the root cause of problems, and even predict problems before they happen. This has been a hard problem to solve. Network telemetry record formats (think broadly here–SNMP, logs, NetFlow, as well as streaming telemetry) were mostly not built with ML in mind. That means for ML to make use of networking data, the data has to be normalized with contextual metadata wrapped around it. I suspect that’s why it’s taken a while for us to see AIOps tools that are truly helpful to network operations come to market. One company that’s gone after the AIOps problem for network engineering is Selector, our sponsor today. We’re talking with Nitin Kumar, co-founder and CTO, as well as engineer-turned-product manager Kevin Kamel. We discuss what Selector is, how it works, and how Selector, as an AIOps tool, can make your life better as a network engineer. Show Links: Selector.ai Selector Analytics Architecture – White Paper PDF Selector Videos – YouTube Selector on LinkedIn Selector on X Selector Copilot: Upcoming Feature Preview – YouTube   Transcript: Please note this transcript is offered as-is without editorial correction from a human. Ethan Banks (00:00:00) – Welcome to Heavy Networking, the flagship podcast of the Packet Pushers podcast Network. Nerding out with you since 2010. And our topic today is one. It is one that upsets me greatly if I’m honest, or at least it has been upsetting to me in the past. Today’s episode, along with a couple of others we’ve released this year, have changed my irritated mind on this particular topic. Now, what topic is that? It is artificial intelligence operations where AI ops AIOps has been making the rounds in networking marketing departments for a few years now, and the big promise has been that AI is going to perform analytical thinking for us and when things are going wrong, make life easier for engineers. And the reality has been somewhat different. We’ve gotten lots of statistical analysis tools branded as AIOps, not really AI, but simpler math models that can detect when certain metrics deviate too far from normal. Fancy threshold alerting was mostly all that was behind the curtain. No machine learning, no artificial intelligence. I am very happy to report that over the last year, we’ve seen AI ops applied to networking in a meaningful way. Ethan Banks (00:01:03) – Tools have come to market that use machine learning and artificial intelligence to analyze vast amounts of telemetry, detect patterns in the noise surface, the root cause of problems,