Orange is planning to create shared Network Operating Centres, based in Poland and Romania, that will service all of its European affiliates.
The company already has a network services centre in Poland that it announced back in 2014, and Leboucher said that the operator wants to “mutualise much more our capacity to create a shared service for our different European countries.”
“Our goal is to manage all of the European country networks from this shared facility. It’s a very important project,” Laurent Leboucher, Group CTO and Senior Vice President Orange Innovation Networks, said.
To enable the project, Leboucher said Orange is investing “a lot of R&D” to introduce “automation, closed loop, improved fault detection for the NOC” for Orange’s fixed and mobile services.
“It is a journey. We are onboarding the different Orange affiliates one by one. But we have good trust from our affiliate community and are creating this shared service.”
Orange’s move to a shared NOC mirrors, to some extent, Deutsche Telekom’s Digital NOC, which the company presented at the recent FutureNet World event. DT’s DNOC provides a centralised capability for alarm analysis across its footprint. It automatically captures 95% of its alarms, and resolves or dispatches 70% of its tickets automatically. The carrier said it has been able to reduce truck rolls by 15%, and it has also been able to do away with the nightshift, with only Tier Three support on call during night time hours.
From hours to minutes
One area where Orange has been proving its automated fault detection capabilities is for mobile voice (VoLTE), where it has been achieving much faster times to discover the causes of poor quality.
“With VoLTE you have defects which may come from the core, from transport, IP, also from the RAN -and when it’s on the RAN you need to know which radio site is impacted.
“Traditionally, it needs about half a day – at least four or five hours – to find the root cause in our service management centre. We achieved the same result in three minutes using the power of Machine Learning time series. So we are able to help people in the SMC to find the problem and to be more proactive.”
Leboucher said that Orange has been looking at automated fault detection “for a long time already” but said that the company had managed to implement it “for real”.
“We started to do it in Romania and our plan is to extend it to other countries.”
The company’s data scientists have created its own algorithms, Leboucher said. It has also been using Google’s public cloud capabilities. “We implemented it also on public cloud. We are working with Google, and our plan is to extend this kind of practice to other services and areas. If it works for time series it can work for fixed network, quality of data services on mobile. There are plenty of places where we can improve using ML and time series.”
Automation focus
As the company introduces increased automation in its network operations, it has two initial areas of focus, Leboucher said. The first is “everything related to change” – so the lifecycle management of a network function using DevSecOps.
Allied to that is monitoring and assurance, correlating assurance with Quality Management and Performance Management to make its NOC smarter.
“We want to reach Level Four, according to TM Forum, by 2025,” Le Boucher said, referring to the TM Forum’s Autonomous Networks Project framework.
“We will focus on areas where we believe there is strong value. So that means especially all the telco workloads, where we consider that agility is important.”
A second focus for network automation is in Global Fixed connectivity. Orange’s backbone is becoming disaggregated “step by step” Leboucher said. “We start with switches and go to P Routers and PE Routers later on.
“So, step by step we wll also make it cloudified. And on top of that, our goal is to implement a platform to onboard services, starting with something that we already do like SD-WAN, but also introducing new services such as SASE services, as well as connectivity to the cloud or public cloud platforms.”
Where service processes are automated, they must also be repeatable. Leboucher gave the example of deploying Private Networks, using a network slicing or hybrid architecture. Each customer will bring different SLA requirements but Orange’s processes must be repeateable and industrial, Leboucher said. “And to be repeatable without automation is almost imposstible.”
Intent-driven intent
An important futher step will be to move to intent-driven automation. “At some point, we will need to reach intent driven automation,” Leboucher said.
“I know it’s a buzzword, but it’s very important. It is how we can move from very basic scripting automation to something which is more abstract but also much more powerful, because it can cope with many diverse situations. We have different affiliates, with diverse sets of configurations. We also have a diverse set of vendors and technologies, so we need to be able to cope with that diversity. This is also why abstraction is important. Intent is important.”
“But I know that moving too intent is hard. And it’s not something that we can just do in a matter of a few months of it will take time.”
Orchestration RFP coming soon
Leboucher also said that the company was developing its work to enable network slicing of its public mobile networks. As part of that it will soon issue an RFP for slice orchestration.
“We’ve deployed the first 5G standalone for a public network in one of our most important European countries – Spain. And then we have a roadmap to move on and implement the other countries and to introduce also all the features to make it end-to-end, because it has also impact on the RAN and all the functions end-to-end to implement the slices.”
“On the IT side we need to have active inventory in real time, you need to have also the orchestration for slices. And of course all the monetisation part. For the monetisation I think we are okay today almost everywhere.
“We still need to implement the orchestration part. We will most certainly issue very soon an RFP to source the right technology. So, it means that we will get to a situation by the end of the year where we have at least the most important brick to implement good features.”
Orange has been using HPE to provide automated orchestration in its Pikeo experimental network, but Leboucher said that shouldn’t necessarily be a steer on who it will select for its production network.
Tooling zone
But there are some features that Orange will bring from Pikeo into production. One is what it calls its “tooling zone.”
“It’s more of a set of tools that we created and built in Pikeo to implement the CI/CD DevSecOps pipeline, including also the security tools, and so on. So it’s a set of tools, mostly based on open source, that today we implement to create what we call the network integration factory.
“The network integration factory is responsible for doing all the activities between planning, unitary test, up to end to end testing, fully automated, and then start a deployment rollout in production, starting in some cases with a small set of areas before rolling out globally.
“So all that process is a pipeline, which is implemented locally, and we’ve built and we’ve created the tool as part of the Pikeo project.”
Standards and open source
As operators move to automated operation of network functions running on cloud platforms, one area of dicussion has been how and whether to standardise the underlying platforms, as well as the automation architecture.
Leboucher sees that the European operator-backed SYLVA and Nephio are both important projects that the company is backing. One specification and standardisation effort that some in the industry previously backed was ONAP.
“ONAP is of course, very interesting. I was very involved from the beginning because I was working with AT&T at the time. Orange was the first operator to join the initiative when ONAP was started some years ago, in 2017. Now, although we consider that ONAP brings a lot as a community, we also consider that there is a need to put efforts on Nephio, which is an expedition project.
“Also it’s about using Kubernetes as the platform to orchestrate not only compute and storage but also networks for operators.
“For orchestration, at some point, I believe that the reference implementation could be Nephio. At least for at least for containerised, telco workloads. It will also be possible to cope with legacy networks, but at the least for containerised, more cloud-native telco workloads. Now SYLVA is of course very important. We believe that there is a need to standardise completely and commoditise the telco cloud stack and this is the goal of SYLVA. We are glad to see that momentum is growing and goes from operator community standpoint, but also with vendors. We still need to increase the momentum. But I think we are on the right path.”