Sprint Tears Up the Rulebook on the Edge

LONDON -- Edge Computing Congress -- Telcos sometimes appear no closer to the "edge" than when the concept was first aired years ago. By taking IT resources normally found in large data centers, and introducing them into smaller facilities dotted about the network, operators should be able to host applications nearer to their customers, at the so-called network edge. This shortening of the distance that a signal must travel would cut costs and give rise to new services, said advocates. Yet the industry is still waiting for a major telco to announce a multi-billion-dollar investment in edge computing.

Sprint has at least figured out how it wants to proceed. Two or three years ago, the fourth-biggest US carrier began reviewing its strategy for tackling the Internet of Things (IoT), a market that operators might better serve with an edge computing program. It came up with three fundamental principles, said Parvin Mahajan, the chief of customer operations for Sprint's fledgling IoT business.

The first was to build a core network separate from the one used for Sprint's mainstream consumer business, said Mahajan during a panel discussion at this week's Edge Computing Congress here in London. Sprint also decided to make edge investments, including the rollout of a distributed core network, at about 150 points of presence. Its third principle was to virtualize everything it could.

In addition, Sprint decided it would have to be a lot more accommodating and software-based than telcos usually appear. Above all, that would mean responding to customer demands for edge nodes on top of the 150 in Sprint's plan and allowing third parties to run their software at Sprint's facilities. "Telcos are typically resistant to that," said Mahajan.

Some industry nerviness is certainly apparent. The UK's BT was asked during a phone call with analysts earlier this year if it would be willing to let players such as Google and Amazon into its edge. At the time, Howard Watson, its chief technology officer, declined to answer. "I think that's something that we spend quite a lot of time thinking about and isn't something we're going to comment on at this stage," he said.

Telco wariness is not altogether surprising. An industry trope is that operators have lost out in the mobile data age to Internet companies, who are piggybacking on telcos' costly network investments to provide their own more highly prized applications. Operators are well positioned in the edge, say their champions, because they own so much of the real estate between the core data center and the end user. But telcos' ownership of last-mile infrastructure did not halt the march of the Internet companies in the mainstream consumer business.

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

An edge battle between telcos and Internet companies is already happening, according to Australia's Telstra. In June, at London's 5G World event, Channa Seneviratne, Telstra's executive director of network and infrastructure engineering, said he had been approached by Internet companies trying to restrict the telco's edge activities. "I won't mention names, but we've already had exclusive deals offered where they want to provide infrastructure and for us to provide only connectivity," he told conference attendees. "We don't see ourselves just providing dumb pipes."

That will be a concern for Sprint as it starts to roll out its 5G network. "Our projection is that 10% of the IoT revenues will be from connectivity and the rest will be from apps and services," said Mahajan. That demands a different approach from the old-fashioned one of buying spectrum, building a network and then searching for customers, he emphasized. "You can't build a network and bring services to it. You have to see where demand is and take the network to that."

As telcos continue to weigh their edge strategies, that change in mindset could be the hardest transformation of all.

Related posts:

— Iain Morris, International Editor, Light Reading