The road to autonomous networks is chock full of AI, ML and closed-loop automation for service providers.
While closed-loop automation within service providers' networks has been ongoing for years now, 5G, multi-cloud, and edge compute use cases require deeper visibility into networks and services. Closed-loop automation takes human error—and in some cases, jobs—out of the equation by eliminating manual tasks and processes while also proactively fixing network issues.
On Friday, Juniper announced its Paragon Automation platform, which it said is a modular portfolio of cloud-native applications that can enable closed-loop automation. Unlike some vendor's automation solutions, Juniper said Paragon was build from the ground up for 5G and multi-cloud environments.
Prior to the advent of closed-loop automation, service providers relied on passive traps in their networks to collect data, or, in some cases, customers reporting network issues after they've already occurred. Service providers, such as Verizon, have been building service chains coupled with closed-loop service assurance to implement business outcomes in the network and not just inboxes with blinking red or green lights.
Machine learning and artificial intelligence are the key elements for building closed-loop automation in today's more complex networks, and, eventually, for autonomous networks.
In an interview with FierceTelecom, Colt Technology Services CEO Keri Gilder said Colt is moving towards a more predictive capability for a better understanding of the performance, monitoring and fault indicators of a circuit. All of which allows Colt to remotely navigate its network and cut down on truck rolls, among other benefits, by investing in fault and performance monitoring.
"Within Colt we have conducted production trials with equipment that is constantly monitoring the network for any event, and then based on AI learns from that enormous data lake to predict when things are starting to deteriorate, allowing us to act before critical faults—like a hardware port becoming— arise, and in the weeks before that shows some innocent but predictive behavior," said Colt's Peter Coppens, vice president of the product portfolio. "This is happening within the Colt domain."
RELATED: Colt tees up proof-of-concept trials on new network edge nodes
Colt is also kicking the tires on end-to-end closed-loop automation with other service providers.
"I actually like to think about it from almost cloud-to-connectivity perspective," Gilder said. "Now, those are separate systems. You have cloud and you have one screen that you have to look up there. And then you have your connectivity and another screen there.
"What we're looking at involves the entire ecosystem. So we can actually start to see within the full circuit, from end-to-end, whether it's on our network or somebody else's network. Obviously, we'll need to have the relationships and understanding for monitoring with our friends and other providers. But if we can come to those agreements, then I think we do have this closed-loop kind of capability where it's literally a 360 degree view of what's going on across the across the entire cloud-to-connectivity profile."
Coppens said Colt has been in discussion with some cloud providers on the possibility of expanding closed-loop automation across domains to provide a real end-to-end view of the performance across networks and clouds.
While end-to-end automation and orchestration efforts are well underway with projects such as ONAP, closed-loop automation between carriers looks like a good fit for MEF and its LSO APIs.
"We are working on trouble ticketing, which if both side use service orchestrations can close the loop using automation," said MEF CTO Pascal Menezes, in an email to FierceTelecom. "I would not call this a fully closed-loop automation system where both sides can self diagnose and fix using machines. Our Interlude APIs have SOAM (Service OAM), which will do this but we are still a ways out for full automation."
BT banks on closed-loop for its digital twin
BT has been building common service models, or model-based concepts, to create product models across its wider IT infrastructure. Automated configuration allows BT to see what parts of the network need to be upgraded, rolled back or left alone when doing network upgrades.
RELATED: BT preps its network cloud for global deployment next year
BT Chief Architect Neil McRae said his company has also built a digital twin of its network and cloud platform. A digital twin is the generation or collection of digital data representing a physical object. BT is using ML and AI to optimize its network and help drive better performance.
"We've experimented with ML and AI in the kind of operations side of our business," he said. "How can we use ML and AI to help us avoid problems and also help us restore things if there is a problem? We're using those types of solutions to really help our engineering teams build networks better
"All the capabilities of both ML and AI are built into our platform. We really believe in the digital twin approach. We think those (ML and AI) capabilities are critical because the network is now so complicated. It's just impossible for humans to do everything."
In order to improve its network operations, McRae said BT has closed-loop automation live in its network and digital twin.
"With Covid and the swing of everyone working from home, that capability was really valuable for us because we're able to rearrange some of the capacity in our network to cope with a change in demand both on our mobile network, but also in our fixed network," he said.
Google drives towards autonomous networks
Last year, Google Cloud announced its telecommunications strategy that included its Anthos for Telecom, which is a platform for delivering workloads to the network edge on Google Cloud. Google Cloud has combined Anthos, AI, and machine learning with its global edge network and its telco partners’ networks to tie 5G into the cloud.
RELATED: Google Cloud, AWS and Azure want to 'friend' telcos, but not every telco is ready to click 'accept'
Google Cloud is keen on bringing Google's internal experience with AI and ML into the service provider realm to enable closed-loop automation.
"Because the service providers have a tremendous amount of telemetry available to them, we say 'We have the AI tools and analytics capabilities for you to drive in real time, or in near real time, insights on that telemetry, and then do the closed-loop,'" said Google Cloud's Amol Phadke, managing director of telecom industry solutions. "You can actually increase network performance or manage your network better or do capacity better.
"We are working with a number of (service provider) partners right now for them to take our closed-loop automation and network analytics into their infrastructure. We are doing this completely in sync with whatever their current plans are around trust, autonomy and security. And Google is coming in saying 'Hey, the telemetry is still yours. We are simply giving you the insights and the algorithms we built in optimizing our networks and putting them in yours.'"
Phadke said Google Cloud was driving deeper integration of AI into service providers' networks, including in their the core networks. AI, ML and closed-loop automation are the stepping-stones to the Holy Grail of fully autonomous networks for service providers and Google.
"That autonomous network is exactly the vision we ourselves are trying to get to for our network," he said. "Google itself has probably one of the largest privately owned networks, and we actually continue to build and scale it to match the demand that Google applications face every day from billions of users.
"That capacity continues to increase, and for us to operate that at scale our internal goals are to actually drive towards an autonomous network."