Nobody Has the Foggiest About the Edge

Edge computing gives data centers little appearance | Knowledge of data centers

Cisco once tried to popularize fog computing as an alternative to the edge. If you’ve never heard of it, you’re not alone. For a brief period, fog computing gave rise to a new industry group called the OpenFog Consortium (because the industry needed another group with an open name prefix), but this was eventually absorbed into a larger association known as the Industry IoT. Consortium. Mentioned zero at Informa’s edge event in London this week, fog computing has evaporated like morning fog.

Too bad, because it’s a better description of the edge than “the edge” and it’s also a metaphor that shows the confusion of the industry in this area. As an expression, the edge doesn’t really make sense at all, suggesting that networks have a solid border like a baseball field. If they do, it’s the coverage limit in rural areas β€” and that’s not the kind of place anyone would necessarily want to put a lot of computing power into. Fog, on the other hand, refers to the drift and dispersal of IT resources across the network – from central data facilities to user gadgets.

No one in the telecom sector is very clear on whether this decentralization makes sense or what form it could take. Operators are generally in favor of centralization. “We don’t want to put a lot of edge clouds,” Terje Jensen, head of network architecture at Telenor Norway, said during a presentation at Informa’s event in London. “We want to centralize as much as possible.”

Providing the landscape with ‘peripheral centres’, even if it means retrofitting existing buildings, probably wouldn’t come cheap. Operators should presumably at the very least invest in additional server equipment and software licenses. Unsurprisingly, then, the cheerleaders for this network re-architecture are hardware vendors such as Intel and software developers, big or small. Intel estimates that the edge silicon market will be worth $65 billion by 2024 (it’s unclear if this represents cumulative industry revenue by then, but Gartner estimated annual semiconductor sales at $595 billion last year).

Puking in the metaverse

There are a few reasons an operator might do it, Jensen admitted. The most exciting thing is all about service improvement. By shortening the journey for a data signal between the customer and the site, an operator can reduce latency, a delay that could disrupt more sophisticated applications. Otherwise, virtual reality would be disorienting enough to make headset wearers puke on their omnidirectional treadmills after just a few seconds in the metaverse.

But the metaverse remains science fiction, and operators don’t want to build a low-latency infrastructure for it if all the extra revenue ends up with Facebook or other members of Big Tech. Today’s latency-sensitive applications do not require an edge site for every city. Three UK believes it can meet customers’ needs for “reduced latency” over a network made up of just three big data centers and 19 edge facilities.

Other UK operators seem to broadly agree. β€œThe fiber lengths are not as long as in the US and the incremental benefits of latency are quite small compared to major countries,” said Viraj Abhayawardhana, who works in network strategy at BT. Vodafone UK, which is partnering with AWS on the edge, expects to make do with a handful of edge sites at most. So far, it has only deployed one, while its German brother has three.

However, business customers may want data processing to be done locally, the BT executive said, at least in part to ensure they comply with data sovereignty laws. But this doesn’t sound like a particularly compelling “use case” for the telco edge. For starters, sovereignty issues are mostly national, no longer localized (any data center in the UK would likely conform to a UK company regardless of location). And a company deeply concerned that data could fall into the wrong hands seems more likely to have its own private cloud. There could still be a possibility for a telco – but it’s less clear.

The other major reason to invest in the edge has to do with internal efficiency. In France, infrastructure owner Cellnex is helping to move parts of a radio access network (RAN) operated by Bouygues to say 100 edge facilities. But this is more about centralization than decentralization. IT resources in the RAN are usually located at the bottom of a mast. A ‘cloud RAN’, as operators call it, means they are sucked into a smaller number of facilities and use less IT equipment. Without suitable fiber connections back to the masts, latency becomes an issue.

For this reason, Tommi Uitto, the head of Nokia’s mobile business, thinks that many operators will only consolidate their RAN exchanges, which are responsible for the functions that are not latency sensitive, and leave the so-called distributed units (DUs) on masts. “The DU should not be more than 20 kilometers from the cell site because it has real-time sensitive features,” he said in a recent interview. “In practice, it might not make sense for someone to build a new grid of these kinds of edge sites, which is why we think most DUs will actually be on the mobile site.”

Non-glamorous assistants

That makes it even more difficult to imagine an edge implementation with hundreds of sites in a given country. As an operator could use those sites for both DU functions and other IT resources, the business case would undoubtedly be stronger. But operators are increasingly looking like Big Tech’s unglamorous assistants on the customer-facing side, clinging on where they can. Amazon doesn’t need a telco partnership to sell AWS to a business or Amazon Prime to a household. Does it need one to sell edge services?

Read the rest of the article at Data Center Knowledge’s sister publication, Light Reading.

Leave a Comment

Your email address will not be published.