Cisco once tried to promote fog computing as an alternative to the edge. If you haven’t heard of it before, you’re not alone. For a short time, fog computing gave rise to a new industry group called the OpenFog Consortium (because the industry needed another group with an open prefix for its name), but this was eventually absorbed into a larger consortium known as the Industrial Internet of Things Consortium. Not once was it mentioned at an Informa event in London this week that fog computing has evaporated like morning mist.
A pity, as it is a better description of edge than “edge,” as well as a metaphor that appropriately captures the industry’s confusion in this area. As a term, the edge doesn’t really make sense at all, suggesting that networks have fixed boundaries like a baseball field. If they do, that’s the limit for coverage in rural areas – that’s not the kind of place where someone might necessarily want to put a lot of computing power. By contrast, fog refers to the skew and dispersion of IT resources across the entire network – from central data facilities to user tools.
Nobody in the field of communications is very clear on whether this decentralization makes sense or what form it can take. Usually, centralized operators are preferred. “We don’t want to put a lot of external clouds in there,” Terje Jensen, head of network engineering at Norway’s Telenor, said during a presentation at the Informa event in London. “We want to focus as much as we can.”
Dotting the landscape with ‘edge centers’, even if it means outfitting existing buildings, probably won’t be cheap. Presumably, operators will have to invest in server equipment and additional software licenses, at least. Unsurprisingly, then, network reengineering clappers include hardware vendors such as Intel and software developers, both large and small. Intel estimates that the “silicon edge” market will be worth $65 billion by 2024 (it’s unclear whether this means cumulative industry revenue by that time, but Gartner estimated annual semiconductor sales at $595 billion last year).
vomiting in metaverse
There are several reasons why the operator would do this, Jensen admitted. The most exciting is all about improving the service. By shortening the data signal journey between the client and the site, the operator can reduce latency, a lag that may interfere with more advanced applications. Otherwise, virtual reality would be confusing enough to make headphone wearers vomit on the multi-directional treadmill after just a few seconds in the metaverse.
But the real meaning remains science fiction, and operators don’t want to build the low-latency infrastructure for it if all the extra revenue ends up with Facebook or other Big Tech members. Today’s latency-sensitive applications do not need a border location for each city. Three UK believes it can meet customers’ needs for ‘low latency’ with a network of just three massive data centers and 19 state-of-the-art facilities.
Other UK operators generally seem to agree. “The fiber lengths are not as long as they are in the US and the additional latency benefits are very small compared to larger countries,” said Viraj Abhayawardhana, who works on network strategy at BT. Vodafone UK, which partners with AWS on Edge, expects to be content with a few edge sites, at most. So far, she has only posted one, while her German brother works three.
However, business customers may wish to conduct data processing locally, according to the CEO of BT, at least in part to ensure they comply with data sovereignty laws. But that doesn’t sound like a particularly compelling “use case” for Edge Communications. For one thing, sovereignty concerns tend to be national, not local (any UK data center would likely be suitable for UK business, regardless of location). And the company, so concerned that data could fall into the wrong hands, seems likely to have its own cloud. There is still a telecom opportunity – but it’s less obvious.
The other main reason to invest in Edge is internal efficiency. In France, infrastructure owner Cellnex is helping to transform parts of the Wireless Access Network (RAN) operated by Bouygues into 100 state-of-the-art facilities, for example. But it is about centralization, not decentralization. IT resources are usually located in the RAN below the mast. “Cloud RAN,” as operators call it, means bringing them to fewer facilities and using less IT equipment. Without proper fiber connections to the masts, latency becomes an issue.
For this reason, Tommi Uitto, head of Nokia’s mobile business, believes that many operators will only integrate their central RAN units, responsible for non-latency-sensitive functions, and leave so-called distributed units (DUs) in the masts. “Depleted uranium should not be more than 20 kilometers from the cell site because it contains real-time sensitive functions,” he said during a recent interview. “In practice, it might not make sense for anyone to build a new network of this type of terminal site, which is why we think most of the depleted uranium will already be at the cell site.”
This makes it difficult to envision a sophisticated publication that reaches hundreds of sites in a given country. If the operator can use these sites for both DU functionality and other IT resources, the feasibility study will undoubtedly be even stronger. But operators are increasingly looking like the unattractive assistants of big tech on the customer-facing side, clinging where they can. Amazon does not require a communications partnership to sell AWS to a business or Amazon Prime to a family. Do you need one to sell advanced services?
Read the rest of the article In the data center sister publication, Light Reading.