Telcos risk paying a heavy price for the cloud
Visitors to last week's Mobile World Congress in Barcelona could flit between most stands without having to change outfit, meet any strict criteria or even don the requisite lanyard. Alas, there is no such freedom of movement for telco software between the clouds now able to host it. Network functions (or NFs) developed for one vendor's cloud cannot enter another's without expensive repurposing by their suppliers. "Then they pass the cost to us," complained Michael Trabbia, the chief technology officer of France's Orange.
One option is simply to maintain multiple clouds for different vendor functions, a separate residence for each supplier that none can ever leave. Economically, the telco is probably no better off, and if the home suffers damage the software has nowhere else to go. An alternative is to house all functions in one of the many-bedroomed mansions of a public cloud. "When you are AWS, the market you propose to Ericsson is bigger, so Ericsson has an interest in doing that," said Trabbia. "But we want to have choice and not be dependent on a hyperscaler solution."
Ericsson agrees there is a problem. "How to deploy software in different cloud environments – at a high level, it is hard making that work in practice," said Per Narvinger, the head of Ericsson's cloud software and services unit. "You have hyperscalers with their offering and groups trying to standardize and people trying to do it their own way. There needs to be harmonization of what is wanted."
Europe's big telcos, together with Ericsson and its Finnish rival Nokia, have teamed up on a standardization initiative dubbed Sylva. The goal is to come up with open source, non-proprietary cloud tools. If enough stakeholders back it, Sylva could feasibly put an end to the cloud silos and even produce an alternative to the hyperscalers. "What really matters is scale, having enough telcos together on that project so that Ericsson and Nokia and others can integrate only once," said Trabbia. "This will be much cheaper."
Right now, the lure of AWS, Google Cloud and Microsoft Azure – the big three aforementioned hyperscalers – sounds hard to resist. On a like-for-like basis, using Azure costs 38% less over a five-year period than relying on what is disparagingly called a "do-it-yourself" or "DIY" cloud, according to a new white paper (likely sponsored by Microsoft) written by Analysys Mason, a market research and consulting firm.
Its authors neatly summarize the chief DIY issue for telcos in their conclusions: "They have accumulated multiple network cloud silos using various tightly integrated xNFs and clouds, and have built specific automations that are not transferrable to other clouds." A recommendation that conveniently ignores concern about hyperscaler dependency is to extend Azure – well established as an option for standard IT workloads – to other cloud domains, including the "existing network cloud silos."
Strictly speaking, there are some network domains that should forever be off-limits to the public cloud. Parts of the radio access network (RAN), in particular, have such demanding performance requirements to do with latency (a measure of the roundtrip time for a data signal on the network) that hosting them in a public cloud facility – a long way from many radio sites – would simply not work. Even Microsoft has acknowledged this shortcoming.
To address it, however, the company has developed an on-premises pitch. Essentially, it argues, the same Azure platforms that would host IT workloads or business support systems in a Microsoft data center can be deployed in a facility owned by a telco and then used to support latency-sensitive network functions. Tommi Uitto, the head of Nokia's mobile network unit, says his company's latest RAN software could sit on this Azure platform just as comfortably as it could on other clouds.
The 38% saving in the Analysys Mason report was modeled precisely according to this on-premises scenario, and Microsoft's platform is ultimately intended for the very "far edge" near radio sites, according to Yousef Khalidi, the corporate vice president of Azure for Operators. What's more, as far as he is concerned, the definition of DIY clouds includes those developed by telcos in partnership with private-cloud platform specialists such as Red Hat, VMware and Wind River.
The Microsoft move, then, somewhat undermines one of the main arguments for retaining one of these specialists in addition to a public cloud. "Hyperscalers are not yet able to go to the far edge and host a real-time radio application," said Paul Miller, Wind River's chief technology officer, during an interview with Light Reading last year. If Khalidi is right, that point of differentiation is disappearing.
Dealing with lock-in
But the likes of Red Hat and VMware can also be seen as potential solutions to the problem of public cloud lock-in. The basic theory is that a telco running software on their platforms inside a public cloud would retain the option of moving these platforms to another cloud. "You can very much think of it as an abstraction layer or point of consistency for things to run on independently of the underlying infrastructure," said Chris Wright, Red Hat's chief technology officer.
It is hard to imagine a Microsoft pitch along these lines, allowing a telco to use some of its systems within a broader AWS or Google Cloud framework. The risk previously identified by James Crawshaw, a principal analyst at Omdia (a sister company to Light Reading), is that Wright's so-called "abstraction layer" simply becomes a different source of dependency. If telcos arrive at this conclusion, they might just decide it is safer and more economical to rely solely on a behemoth like Microsoft – with its $1.9 trillion valuation and array of services –than involve a relative minnow and one-trick pony such as VMware, currently worth about $51 billion on the New York Stock Exchange.
Not all telco executives are convinced the public cloud is necessarily cheaper than the alternatives. "I don't think it is proven yet," said Howard Watson, BT's chief security and networks officer. "They are running these cloud environments at enormous scale … but the economics don't yet reflect that." Outside the telecom sector, others have insisted the economic attractions of the public cloud diminish as a customer grows and its IT usage becomes more predictable.
Such considerations, along with the fear of being tied to one US tech giant's offer, partly explain why few telcos outside the US have rushed to embrace the public cloud. For its own telco cloud, BT uses components from Cisco and Dell, an OpenStack platform built by Canonical and orchestration tools from Juniper. This seems exactly the type of antiquated and expensive DIY offer that Microsoft derides, brandishing Analysys Mason's new report as backup.
Many technology executives will be more easily persuaded than Watson is. But if operators more broadly have learned anything in the last few years, it is probably that single-supplier relationships and lock-in can be the biggest costs of all.
- Sylva marks latest telco backlash against US public cloud
- Orange confident about open RAN's role in network transformation
- Did AWS win MWC23?
- MWC recap: Google, Microsoft and Oracle diverge on network functions
- Google Cloud takes a non-threatening stance in pursuit of telecom
— Iain Morris, International Editor, Light Reading