Let’s say that you wanted to justify building a bridge between Lubec, Maine and Canada’s Westport, Nova Scotia, crossing the Bay of Fundy. It would be fair to say that such a bridge would be enabling for those who wanted to drive between the points quickly. Our new bridge would be about 50 miles long. The current road route would be about 500 miles. Think of all the driving time that bridge would save (probably ten hours or more)! And if there were a bridge couldn’t you expect people to walk on it? An experienced walker could do that hike in a day…maybe. So we could use the driving and walking benefits of the Lubec-Westport Bridge to justify its surely-enormous cost, right?
The answer to that has a lot to do with how we justify things like 5G and edge computing.
If we had such a bridge, perhaps some would walk on it. Perhaps many would drive on it, but chances are that the numbers who would use the bridge would take centuries to justify its cost. Thus, there are things that I can use the bridge, or a technology, for that would not justify it. Exploiting a resource is one thing, financing it is another.
OK, let’s take this a little further. Grand Manan Island is maybe 15 miles along the path of our bridge. Suppose we build a smaller bridge just that far? It could be a step along the way, right? Yes, but if there aren’t a lot of people trying to drive or walk between Lubec and Westport, there are surely a lot less trying to drive/walk between Lubec and Grand Manan. The only value of that little step would be the value it presented in taking the next longer one. Even adding additional (non-existent) islands to the route wouldn’t help; no additional island would likely contribute much to the business case, and if any step were deemed to be unjustified, the whole value proposition would be shot.
By this time, I expect that you can see what I (at least) believe the connections are between my bridge analogy and things like 5G or edge computing. We have no problem deciding what we could do with either, but we’re conspicuously short of things that could justify them.
When I first ran my models on “carrier cloud”, I identified six theoretical drivers: virtual CPE and NFV, advertising and content, 5G, public cloud services offered by operators, contextual services, and IoT. My model suggested that NFV and vCPE had minimal potential. It also said that 5G would have its greatest potential impact by 2022, and that in the longer term, IoT was the decisive driver. In other words, my model said that we could visualize six islands between Maine and Nova Scotia, and that each of them could (if properly exploited) contribute to an investment that would then be exploited by subsequent service steps.
What happens to this picture of “successive evolution” if one or more steps doesn’t play its role? In the case of edge computing (carrier cloud), the answer is that the operators never make any major edge investment. The first of the possible drivers, NFV, never really had much of a chance except in the context of 5G, and operators have been increasingly looking to the cloud providers for hosting 5G virtual functions. Operators never deployed their own video ad and caching services to any extent, and that rounds out all the early edge applications.
Contextual services and IoT are related; the former relies on the latter to get real-world information on a service user, and presents AR/VR or other augmented information forms based on the combination of data available. Because of their early misjudging of the 5G opportunity (they wanted to charge “things” in IoT for cell service when they ran out of humans), operators have done nothing useful in the last two of our drivers so far, and time there is running out.
Neither 5G nor edge computing will fail because of operator misperceptions. 5G is the technical successor to 4G, which was the successor to 3G. There was never a risk it wouldn’t deploy, just a risk that it wouldn’t generate any incremental revenue. Edge computing, on the other hand, could be at risk, and with it a whole lot of vendor revenues.
My model said a decade ago that were operators to deploy edge computing (carrier cloud) at the scale all the drivers could justify, it would be the largest single source of new data center deployments in the history of IT, with the potential for one hundred thousand new edge hosting points. A big part of that number comes from “competitive overbuild”, where multiple operators locate edge hosting sites in the same area because they’re competing for the services those sites support. If, as it appears will be the case, the public cloud providers end up deploying all the edge computing, there are fewer competitors, fewer data centers, and less vendor revenue to fill those centers with network and computing gear.
This is why vendors should be working hard to devise a strategy for edge computing that operators could buy into. That strategy would obviously have to be centered in the metro zone, because metro opportunity density is high enough to justify deployments and close enough to the user to still present attractive latency options to applications.
There are two credible edge opportunities, IoT and metaverse hosting, and both have that perfect combination of great potential and technical and business hurdles that seem to characterize tech these days. There are some things that could be done to promote both these applications in a general way, and as I noted in earlier blogs, we could define a “metaverse of things” model that could further harness common elements. And, of course, we could let nature take its course on both. Which option would offer the best, the fastest?
I’m skeptical that IoT applications will drive edge hosting spontaneously, because the specific kind of IoT that would be most specifically an edge hosting application would also require the highest “first cost” of deployment. Enterprises already know that their own IoT needs tend to be tied to fixed locations, like warehouses or factories. This kind of IoT cries out for local, customer-owned, edge computing, because hosting further from the actual IoT elements only increases latency and magnifies the risk of loss of service.
“Roaming IoT”, characterized by the transportation vertical’s use on shipments and vehicles, ships, aircraft, and trains, is different because it doesn’t have a fixed point to place an edge, and in fact is necessarily highly mobile. I’ve worked on this vertical for decades, and I can see IoT edge hosting options, but they would be more likely to exploit edge computing than justify it, particularly because you’d need edge resources through the entire scope of movement of your IoT elements.
Metaverse hosting is in a sense the opposite; there are strong reasons to say that it depends on edge computing, but a lot of fuzziness on the issue of just how it makes money or even works at the technical level. If we presumed that a “metaverse” was a social-network foundation, then social-network providers (like Meta) would surely tune it to fit what they’d be prepared to deploy. The opportunity in edge computing depends on the presumption that there would be a lot of metaverses to host, making the business of hosting them a good one.
Given that one headline this week was that Walmart plans to enter the metaverse, you’d think that we were already on the verge of a metaverse explosion. Not so; we’re on the verge of another “label-something-to-create-click-bait” wave. What Walmart is actually contemplating is creating a cryptocurrency and NFTs, neither of which can be said to mandate edge computing, and which are in fact more likely aligned with the Web3 stuff. As I noted in a prior blog, Web3 is also mostly hype, but it does suggest that some form of digital wallet and some mechanism for NFT administration could be overlaid on current Internet technology, particularly by players who take payments.
We’ve had credit-card terminals for half a century or more and somehow managed to support them without edge computing. Adding security through blockchain is a worthy goal, but it doesn’t require that we host anything at or near the edge, because credit card and other retail transactions are done at human speed and can tolerate latencies measured in single-digit seconds.
I think that the metaverse may well be the key to edge success, but only if it develops across its full range of potential applications. It’s too early to say whether that will happen, but I’ll blog on the range of “potential applications” to set the stage, and watch for indications that multiple options are moving to realization. If we see that, then edge computing will become a reality, likely even in 2022.