Action Week 2020: Building the autonomous network with Lego blocks

Automation is the only way forward for communications service providers (CSPs) – the volume and velocity of changes in 5G networks demand it. The question is: How to do it at a reasonable cost?

In his keynote at TM Forum Action Week Tuesday, Huawei’s Kevin McDonnell, Senior Director, Intelligent Automation, Ireland Research Centre, explained how TM Forum members are addressing the challenges around automating network operations through the Autonomous Networks Project. Using Lego blocks as an analogy, he described how CSPs can use intent-based management and control loops to dramatically simplify their networks and operations and achieve automation.

Huawei's Kevin McDonnell

Huawei’s Kevin McDonnell

“The genius of Lego wasn’t the four dots on top,” McDonnell said. “The genius of Lego was the bottom of the block because that’s how it gets backward compatibility. If you take any Lego set, you can snap the blocks together because of that.”

The Autonomous Networks Project envisions using intent-based management combined with application program interfaces (APIs) acting as Lego blocks to connect different operational “domains”. In intent-based management, domains are used to abstract the complexity of the network at a high level and expose network services (service function chains made up of interconnected virtual and/or physical network functions) through a standard API. So, in autonomous networks, TM Forum Open APIs can act as Lego blocks.

“The key is that it is fully automated; it’s self-healing and self-optimizing,” McDonnell said. “A self-healing network can fix itself; self-optimizing means it’s using intelligence to bring that automation.”

This happens across all domains in a CSP’s environment – from resource management to operations and maintenance and from customer experience to service enablement.

McDonnell said TM Forum is taking “a clever approach” to autonomous networks for three reasons:

    1. The project is holistic, meaning it is taking into consideration the work of other standards-development organizations and open source groups such as ONAP (the Open Network Automation Platform).
    2. The team is using an Agile approach, testing ideas in Catalyst proofs of concept – project members are participating in 12 different projects.
    3. Most importantly, the approach focuses on ecosystems that include other verticals such as the automotive industry, which is at least double the size of the telecoms industry.

What’s an Autonomous Domain?

McDonnell explained the project’s concepts further, introducing the idea of an Autonomous Domain to simplify management and orchestration. “An autonomous network is a network that can measure, analyze and control itself independently,” he said. “A domain is something similar but it’s an abstraction. The Autonomous Domain is operations and networking elements together, but the key point is that it’s a well-defined boundary.”

The Autonomous Domain operates under a set of rules. For example, it has a single control loop and is exposed externally through APIs.

The Vegas Rule

Another important rule is what McDonnell jokingly called “the Vegas Rule”.

“What happens in the domain stays in the domain,” he explained. “No matter how badly it goes – whatever happens – we keep it in the domain.

Domains can be within a CSP’s environment or include a partner’s. “You can have domains that go up through the stack with additional operational services, such as perhaps SD-WAN or VoLTE,” McDonnell said. “So, there is flexibility in how they’re developed.”

The ‘cross layer loop’

Finally, the idea of a “cross layer loop” is important to the team’s work. This is not a control loop, rather it’s a “learning” loop. McDonnell compared it to a set of interlocking gears: “Business automation and network automation work together to create this cross layer loop.”

The Autonomous Networks team will have to work with the team that is developing the Open Digital Architecture (ODA) to advance this concept.

“That cross layer loop lies in the intelligence management part of ODA, so we will need to work with [them] to figure out how to actually do this,” McDonnell explained.

In conclusion, he said he’s optimistic the Autonomous Networks Project will be successful.

“It’s a very ambitious project, but I think that we can do something valuable with it,” he said. “There won’t be a magic breakthrough – it’s really a series of incremental improvements that get you there. We all believe that we need these new networks that are easier to build, manage and operate.”

Leave a Reply