The Self-Driving Network™ Episode 4_ The Progression copy

The vision of a self-driving network might seem a very difficult one. And many people ask me, "Is that a ten-year journey? Is that a five-year journey? How do we get there?" The first thing is there's some base technologies that you need. Definitely you need automation. Automation makes a lot of the mundane tasks easier to do. . . The second thing you need is telemetry. You need to know what the network is doing. You need the network topology. . . You need to know what kind of flaws are going on, who's talking to whom. And you need that information in real time, in a way that's consumable by machines so that you can do the analysis. And the next step is that analysis that you want to do. So you want to be able to look at all this data, and then figure out what should be done. And you could use rule-based systems to do that or you could use machine learning to do that. Probably you'd use a combination of those two. And that will get you to an action that you need to take. And then you take that action again going back to the automation, you can push that action out to many things. And if I see something anomalous going on in some part of the network, I can zoom in and say "What's going on there?" and say "Oh, it's okay. It's not bad." Or maybe I have to take some action. And then I'll zoom out again. And so I can have the right level of telemetry whenever I need it. An important component of self-driving networks is intent-driven networking, or declarative programming, where you express what you want the network to do rather than telling every device how to do it. And that makes your job much easier... This is a journey that's going to take a while. But we have a lot of elements today.