

When it comes to closed-loop automation of large-scale networks, it can seem like we’re living in the movie “Groundhog Day.”
For years, we’ve been told that fully automated, self-driving networks were just around the corner. Any day now, the story goes, you’ll be able to provision and activate new services with the push of a button. Your network will detect and correct problems as they happen, all on its own. And yet, when we look at real-world networks, even the most advanced operators have managed only baby steps toward this goal.
What’s the holdup? Is closed-loop automation a myth, an illusion that stays just over the horizon but never quite arrives? You could be forgiven for thinking so. But I’d argue the innovations of the past few years have finally put us on the path to achieving it. To get the rest of the way through, we need to think differently — and more pragmatically — about our approach.
There is no magic solution for closed-loop automation. Modern networks are just too complex, with too many moving parts, to automate everything in one fell swoop. What we can do, however, is take incremental steps toward that goal.
Among the most important: implementing “zero-touch” assurance. Today, we finally have the tools required to add fully automated testing and monitoring of the end-to-end delivery chain for every customer-facing service. As we do, operators and their customers will realize immediate benefits. But they’ll also be putting in place a core enabler for tomorrow’s closed-loop automation.
As the industry pushes toward closed-loop automation, we see lots of emphasis on the concept of zero-touch operations. But what does that actually mean? The self-driving networks of the future will automatically provision and configure equipment to activate a new service. Then, they’ll monitor the service and continually fine-tune it in response to changing conditions, all in a closed-loop fashion.
The industry has made major strides in the first part of that continuum. Large-scale operators now use zero-touch provisioning every day. But in order to get to full closed-loop automation though, we need to nail the monitoring part too — and that’s where we continue to struggle.
Current monitoring approaches infer too much and measure too little. For decades, our approach to monitoring has been to collect metrics from resources along the service delivery path to try to ascertain the health of the service. But it’s inherently limited because it’s based on a construct of the service assembled from management-plane statistics rather than measuring the actual service.
And there will always be differences between models of service and the service itself. It’s also extremely difficult to maintain accurate models in dynamic networks. As we move to tomorrow’s fully virtualized cloud-native networks, where slices extend across multiple domains and clouds, that kind of modeling becomes impossible.
To get to true closed-loop automation, especially as services incorporate more third-party networks and clouds, we need to be able to analyze when the customer experience is bad and identify what interventions are needed to fix it through the end-to-end service delivery chain.
There’s a solution to this conundrum, one that provides the kind of monitoring needed for tomorrow’s closed-loop automation, while improving service quality and reliability right now. It’s automated active testing and monitoring, sometimes called active service assurance or zero-touch active testing and monitoring.
Active testing and monitoring add an extra step to the service activation process — arguably, one operators should already be taking — to validate that the service is working as it should. But the key word here is “active.” This testing doesn’t rely on passive management plane metrics or abstract models of a service. Rather, it runs synthetic transactions right in the data plane along the end-to-end service path.
In this way, it measures the service’s real-world performance instead of inferring it. Once active testing is in place, it can automatically re-run itself after any and every change in the network to validate that the service is still healthy.
By adding zero-touch active assurance to the turn-up process, operators can measure service quality on the data plane, instead of inferring it. They can maintain a more accurate view of the end-to-end service delivery chain. They eliminate the need to maintain more and more elaborate models as networks grow more dynamic and complex. And they take major steps toward self-driving networks, since orchestrators can now measure the real-time health of every service, quickly identify the root cause of problems and identify exactly what’s needed to fix them.
Sounds like a pretty big step toward closed-loop automation, doesn’t it? And because operators have made such huge strides with zero-touch provisioning, adding zero-touch active assurance is now relatively trivial to implement. It just entails running some software and connecting some application programming interfaces.
It may take the industry a few more years to get to full closed-loop automation, but unlike in the past, we can now see the outlines of true self-driving networks finally coming into focus. Active assurance goes a long way in bridging the industry’s gap that remains.
The best part though is that active assurance benefits operators right now. They can ensure services are working as they’re supposed to before handing them off to users, instead of after. And if there’s a problem affecting a service, they can quickly identify and fix it.
The bottom line: With active assurance, operators can detect problems proactively, instead of waiting for customers to find them first and implement a critical component of their self-driving networks framework, killing two birds with one stone.
Julius Francis is head of product marketing and strategy at Juniper Networks Inc. He wrote this article for SiliconANGLE.
THANK YOU