Laurent Leboucher talks to Annie Turner about automation’s pivotal role in network evolution and service innovation
Laurent Leboucher is Group CTO at Orange and leads Orange Innovation Networks. Automation has existed for many years as a way to simplify human operations and this is particularly true over the last few years as telecoms networks have become increasingly complex.
More recently, “with our data and AI experts, we have worked to optimise, automate and avoid many repetitive technical tasks with different use cases,” Leboucher says. He cites using machine learning to track and fix faults in voice over IMS and LTE as a favourite example.
Problems can occur on the transport side, in the core or the RAN and it used to take five to six hours, on average, to identify the root cause, and when it’s in the RAN, find the exact radio site. Applying machine learning to data drawn from across the network has reduced this to minutes, and some of them can be fixed automatically.
“This is extremely powerful, and gives us good hope to extend these techniques to many other use cases. Our ‘AI empowered network’ programme’s main goal is to reap the benefits of using network data to become much more predictive,” he explains. “Even if we are not completely predictive, we can be much smarter in the way we seek out root causes from all the alarms”.
“We want to move to a smart network operating centre that needs no lighting as it doesn’t need people to watch alarms that are handled in automation. We need smart people to handle complex cases who can focus on the issues that need to be solved. Our new job is to design how to automate the resolution.”
Leboucher continues, “Around operations, we have made significant efforts in the past few years sharing NOC – network operating centre facilities both in Europe and Africa. t’s not just a question of monitoring from a central place but how we interface central monitoring and the field operations within the different countries, allowing them to become more efficient.”
Different as well as better
Moving network functions, network services and workloads to the cloud is another critical evolutionary step for networks, and an industry-wide phenomenon, Leboucher says: “This is extremely important and not just about disaggregation; it’s how we move to a new industrial model. We need to apply DevSecOps to network workloads, to make savings and gain the agility to provide new kinds of network services, in particular for Network-as-a-Service [NaaS], mobile private networks and slicing.
Leboucher considers NaaS first for “fixed global connectivity for companies across the world. We have a large international backbone that we continue to invest in, transforming it step by step into a very powerful platform.
This will allow us to provide customers with network on-demand connectivity – including Ethernet, MPLS and SD-WAN – and combine it with additional functions like cloud connectivity and security including SASE [secure access service edge]. We can also extend it to IoT and other added value functions.
“We strongly believe that this is a very important enabler for revenue growth and we need agility to bring in new partners and to enable customers to use digitally this network capability. That’s one part.”
The other part is deploying mobile networks’ new capabilities on demand, such as network slicing within country markets. The slices will be delivered for private mobile networks, “at large scale and in significant numbers, on the same infrastructure.” he says.
“We are working on the deployment of 5G Standalone in all European countries; the first will go live very soon. We need to automate the lifecycle of network functions, to have the same practices on the network side as we use for digital IT. We need vendors to be part of the same pipeline.”
One homogenous pipeline
Orange aims to automate all the downstream activities via the CI/CD pipeline, including security requirements which “we need to fulfil across the whole chain. This means there is a considerable impact on processes and people: it’s not something we will manage in a few weeks,” Leboucher acknowledges.
To help with this important transition, Orange has applied TM Forum’s autonomous network maturity model in several countries.. “It was extremely useful and produced a lot of information, so we want to extend it and use it as a way to monitor progress of that transformation,” he says.
Leboucher continues, “Our long-term vision is a homogenous telco cloud but today that is not yet the case.” The main issue is some vendors insist on vertical mode operations, delivering workloads into their established distinct cloud infrastructures, even when there are several from the same vendor, they do this for each function, resulting in too many clouds operating side by side.
He believes network vendors cannot stay in this vertical mode in the long run because it creates friction and slows down the end-to-end GitOps pipeline. “The telco cloud needs to be shared for all functions in the same network service,” he says, “and there are two ways of doing that.”
“One way is to work with hyperscalers, which will become even more relevant to the market for some use cases especially when national sovereignty is not required. The other way is for operators to come up with a common reference implementation and this is what we do with the Linux Foundation Europe through Project Sylva.”
The project was launched in November 2022: Deutsche Telekom, Ericsson, Nokia, Orange, Telecom Italia, Telefonica and Vodafone are to create an open source, cloud-based, software framework to lessen infrastructure fragmentation in Europe.
Leboucher notes, “We are bringing as many operators and vendors [as we can] to join this initiative because we don’t only want to rely only on hyperscalers; we want to keep the choice and the capacity to do it ourselves.”
For some B2B use cases, Orange may work with hyperscalers to deliver connectivity plus IT functions, “but we need clear accountabilities,” Leboucher says. “We want to keep our core business as an integrator and operator of networks”.
Getting GitOps
Orange also wants to apply GitOps to “the distributed telco cloud, the infrastructure itself and, importantly, the workload to move to something much more intent driven,” he explains. “The beauty of GitOps and the Kubernetes style is defining what you need in a declarative way then the underlying mechanism automates everything by reconciliation with the desired state.”
Kubernetes’ original use was to containerise resources like CPU and storage, so work is needed to expand the model to network resources and apply it in a more holistic way. Linux Foundation’s Nephio project, announced in April 2022, is working on this with Google Cloud and Orange is fully involved.
Leboucher states, “Currently we are focused on how to containerise network functions; moving to cloud network functions, CNFs, is becoming a reality. The new network functions, mostly for the core, are made with containers. It’s not yet completely cloud native, but going in that direction and getting much better than a few years ago. We are really optimistic.
“Next, to get all the benefits, we need to implement the full end-to-end GitOps pipeline. If we learnt that DevOps was the operating model of the cloud, GitOps will be the one for cloud native, connecting Day0 to Day2, shifting from CI/CD (Continuous Integration / Continuous Deployment) to CI/CD/CO (Continuous Integration/Continuous Deployment/Continuous Operation).
“Having this in mind in terms of coverage, applied to network applications, infrastructures, security or even energy consumption, and including the native resiliency and auto/self-healing of the Kubernetes platform, will lead to new operating ways and models that will make us more productive and competitive.”
Network Integration factories
Orange is creating network integration factories to onboard each function individually then orchestrate the full network service, integrating from end-to-end then automating all integrations so that every time there is an update – maybe something that needs a security patch for instance – the process is automated and can be deployed in minutes.
Leboucher continues, “We start with network services which can get immediate agility benefits from becoming truly cloud native and network functions not too complex to begin with: typically routing and switching functions on the backbone, overlay functions such as SD-WAN, signalling, IMS, EPC and extending progressively to new 5G core.
“They must be virtualised or preferably containerised but also become part of a completely automated delivery process via the CI/CD pipeline. We will extend it later to RAN and fixed access network.”
Open RAN, cloud RAN
How does Open RAN fit into this? Leboucher says, “We expect to move in this direction with Open RAN, which includes cloud RAN, but that will take more time. We will start small scale very soon then introduce it more globally. In the end, the pipeline will include network services with core functions and access functions, at the same time, in the same pipeline.
“With Open RAN you have to integrate components from several vendors We are on a learning curve in an industry today which is going pretty fast. We’ll be showing some concrete developments very, very soon.”
Orange will initially deploy Open RAN commercially in Europe at small scale to understand how the model compares with the traditional one. “Then we will increase automation and extend it to more complex sites with massive MIMO.” Leboucher says.
The cloud element of Open RAN is just one other step in the automation journey: “We will also use the RIC, the radio intelligent controller, for automation benefits,” he adds. “Just think of it as the new way to do C-SON [centralised self-organising networks] closer to real time, in less than a second for very innovative use cases”. Orange launched C-SON in more than 20 countries several years ago, automating radio configuration via a 30mn closed loop.
This article was commissioned by and originally appeared on the FutureNet World website, and is reproduced here by kind permission.