If nothing else, the rise and fall of Huawei in parts of Europe has shown that switching from one radio supplier to another is no telco mission impossible.

Costly and disruptive? Quite possibly. But the Huawei bullet train is departing some European telco platforms as quickly as it first arrived.

Shifting from one public cloud to another might not be so easy. Snap, the provider of the Snapchat messaging service, warned in a 2019 regulatory filing that moving systems between public clouds would be "difficult to implement" and demand "significant time and expense."

That hasn't discouraged AT&T, which is loading all its 5G network operations into Microsoft Azure. Any future breakup threatens to be a messy affair.

But fear of being chained to one provider is apparent elsewhere.

"Some years ago, everyone was saying we would have vendor lock-in with Ericsson, Huawei and Nokia and no one mentioned Oracle and Cisco and now the light is on hyperscalers," says Yves Bellego, the director of network strategy for Orange.

"In fact, that risk is something we have always been very concerned about."

Cautiously does it

The French operator is proceeding cautiously through a strategic partnership it announced with Google Cloud in July last year.

Drawn to Google's expertise in data, artificial intelligence (AI) and edge computing, it has been using its cloud as the AI platform during 5G trials.

Pikeo, the "digital twin" test network Orange is building in the northern town of Lannion, also counts Google as the public cloud platform for AI applications.

Yet keeping technology alternatives in the frame is clearly high on Bellego's agenda.

"We do have a philosophy that we want to build our networks from bricks coming from different suppliers and want the capability to switch from one to another," he says.

"The role of the hyperscalers today in the networks remains, in fact, very limited, whether we are talking about telco cloud or AI platforms."

The difficulty of moving from one public cloud to another would depend on what, exactly, an operator has purchased, according to James Crawshaw, a principal analyst with Omdia (a sister company to Light Reading).

"If it is just infrastructure-as-a-service, it is easy to go across," he says. "If you use platform-as-a-service, like the Google Cloud Spanner database, that is not available on AWS and Microsoft Azure."

Want to know more about 5G? Check out our dedicated 5G content channel here on Light Reading.

These bespoke digital services are creating a "stickiness" that makes changing providers difficult, says Crawshaw.

"The Amazon version of Kubernetes [an open-source platform] is different from Azure's and changing would require rework," he explains. "But it is not impossible."

Other experts have noted similar issues. "Amazon's AI solution is different from Microsoft's and all the software is designed to work with AWS," says John Strand, the CEO of Danish advisory firm Strand Consult. "The migration cost is gigantic."

Risk mitigation at Orange seems partly to mean avoiding over-reliance on Google for AI. Besides continuing to invest in its own AI capabilities, Orange will still use AI tools developed by more traditional suppliers, such as Ericsson, says Bellego.

"The interest we have and that Ericsson has is much more in the algorithms and the data than in the platform itself," he says.

Edging forward

The update came as Orange and Google Cloud opened a new lab in Chatillon (near Paris) where companies and app developers will be able to carry out tests of edge computing services.

It features an edge node based on Google's technology as well as 5G infrastructure supplied by Orange. The aim is largely to help cultivate an edge computing market and tackle any security problems or other challenges ahead of a commercial launch.

Orange has not yet made any firm decisions about that launch, including the extent to which it will rely on Google. The hyperscaler's appeal stems largely from its huge ecosystem of app developers already writing for the public cloud.

For the developers that work in that ecosystem, moving to an edge node supported by Google should be quite easy, says Bellego. Orange, then, can focus on how to integrate these services into its network.

Figure 1: Orange hopes a new edge lab will be a place for collaboration with industry.

Its cautiousness has arguably left Orange trailing other big service providers on the rollout of edge computing.

Vodafone UK, notably, already claims to have slashed latency – a measure of the signal delay that occurs on data networks – from more than 50 to just 10 milliseconds through a deployment this year with AWS and its Wavelength technology. Verizon boasts a similar edge service in the US.

But viable use cases are in short number, and the Orange view is that more are needed to justify investment.

"It is pretty difficult to find a business case for a single edge computing service," says Bellego.

"That is similar to virtualization – a single virtualization project could not justify the investment in infrastructure."

Orange is weighing the need for a global plan that would encompass telco cloud, edge computing and open RAN – an overhaul of radio network design and architecture that operators hope will spur competition and lower costs.

With the overlaps between these related initiatives, a multi-country project could massively improve the economics.

Up for discussion within Orange will be how many sites it will need altogether. In a country the size of France, Bellego thinks Orange would probably need tens of sites for the telco cloud and hundreds for the edge and open RAN, where nodes would be deployed tens of kilometers from a basestation.

It is partly by taking the compute functions normally hosted at the basestation sites and consolidating them in these nodes that operators believe they can reduce costs.

All this will demand substantial investments in fiber to connect sites and nodes. Orange can already boast one of the most extensive fiber networks in Western Europe, but that rollout has been a drain on resources, upsetting some shareholders.

More recently, capital expenditure has fallen, down from €7.4 billion (US$8.7 billion) in 2018 to €7.1 billion ($8.3 billion) last year, as free cash flow has stabilized at about €5.5 billion ($6.4 billion).

A spending uptick without the promise of sales growth will not be what investors want to see.

Related posts:

— Iain Morris, International Editor, Light Reading