We may be seeing some clarity in the edge space, arising perhaps surprisingly from some increased confusion. There are more players emerging, more theories promulgating, and more value propositions being tested, to be sure. What’s interesting is that some trends are emerging.
The biggest question in edge computing is just who has the edge, so to speak. That’s a two-dimensional question, too. The first dimension, taking the “has the edge” to mean “has the lead”, reflects the fact that there’s likely to be a dominating set of players, as there is in any market. The second, taking the “has the edge” to mean “owns the edge elements”, reflects the fact that edge dominance requires having something out there to do the dominating.
Let’s start with potential players/winners. In the past, many (myself included) believed that network operators were going to “have” the edge in both senses of the term. Operators have control over edge offices where access networks terminate, they have a mission that could site gear there in 5G, and they have a set of standards and plans (including 5G) that involve function hosting. Most of the hundred-thousand new data centers that I’ve said carrier cloud could deploy were associated with edge hosting.
Another logical set of winners are the public cloud providers. The biggest challenge with edge computing is what operators call “first cost”, the portion of the cash-flow curve after a deployment starts where investment far outstrips revenue and the net cash flow dips negative. First-cost constraints suggest that many edge applications would try to get off to a slow start, but that’s a problem because if there’s a clear edge value, it’s lower latency. You can thus only start slow by starting in a very narrow geography. But cloud providers are hosting resources for the masses, and they might be able to deploy first simply by offering edge-as-a-service to all the other constituencies.
There’s even a case to be made for server/device vendors. Many proposed edge applications relate to IoT, which in turn relates to some form of industrial automation. The companies that have the industrial processes to automate are logically the edge themselves, and they could in theory deploy technology between IoT elements and their computing resources to provide edge processing. Vendors would have to not only support but promote this, creating a framework that would raise early benefits to the users, and lower risks at the same time.
What do any aspiring edge-winners need? Any form of edge computing needs a mission to serve, and mission development isn’t the strong suit of anyone these days. Market education is a tedious process, and over the last decade the notion of an educational sell has become increasingly oxymoronic. One salesperson told me, in 2016, that “if you have to educate, you’ll never make the sale at all.” People don’t earn commissions by not selling.
This gave the early mover advantage, potentially, to the network operators. They had six potential driver applications for carrier cloud, and there were some early-acting ones that could have created an internal mission for deployment. Feature or function hosting is one example. The operators did commit to these missions with things like NFV, but the commitment never developed into significant deployment. In fact, it’s recently become clear that operators would love to outsource their carrier cloud opportunities to the public cloud providers.
Who of course are our second potential winner group. All of the public cloud providers would love to provide edge computing as a service, and all have shown interest. Microsoft and Google seem to be preferred partners in the operators’ view, largely because the operators fear Amazon’s might in the cloud would defeat their own (delayed, half-hearted) efforts to shift to self-deployed cloud assets down the line. The public cloud gang were successful first in tapping off business cloud opportunities from operators who wanted to provide enterprises with cloud services but didn’t want to build out to get them. What disappointed the cloud providers in these early deals was the fact that the operators were largely interested in erecting big attractive cloud-promoting billboards and seeing what rolled in. Not much did, because (as Verizon learned early on), operators aren’t very effective as cloud computing sales organizations.
The cloud providers also realized they had their own first-cost issue. Remember, few public cloud providers own real estate in every major community the way that the operators do. They can hardly run out and buy some, fill it with servers, and hope operators (who had their own facilities all along) would fill the cloud provider edge instead of their own. In fact, cloud providers had edge opportunities independent of whatever missions the telcos might have sent them, and had them several years back. The cloud provider strategy for the edge has focused on cloud-allied appliances.
Instead of buying real estate, or perhaps even equipment, why not supply users with a software stack and perhaps a hardware specification, and let them roll their own, but one integrated with the public cloud provider’s own facilities and software? Things like Amazon’s Greengrass came along as a result of this sort of thinking.
Microsoft has perhaps advanced this to the state of the art in its Azure edge stuff, which promotes the “intelligent edge” as a piece of the “intelligent cloud”. The basic concept is to make edge a subset of the cloud, one that can be a subset or zone of cloud resources, or an independent device. Things like Azure RTOS and Azure IoT Edge create that private-edge-device-to-Azure-cloud linkage, and there’s a software suite that’s more application-centric (IoT, of course) that facilitates implementation, and thus commitment.
Microsoft has also jumped into the telco-cloud mission with a set of “Edge Zone” offerings. These include both 5G services to operators who want to outsource some 5G hosted-function pieces, to Private Edge Zones for enterprise private-5G deployment. The current 5G-centric stuff includes a bunch of application partners who combine to create a complete solution. This is almost certainly intended to be a prototype for edge-computing strategies in the future, not only for possible carrier cloud mission-stealing, but also for general edge computing, especially IoT.
The big advantage of this approach is that it addresses that educational-sell problem. Selling a solution to somebody’s problem is easy. Selling a technology and then educating the buyer to solve their own problem is a lot harder. That suggests that if vendors want to sell edge equipment to users, competing with as-a-service edge models, they’ll need to build application alliances too.
Which, for vendors, is more difficult. The problem is that while users welcome as-a-service strategies because they’re inherently immune to stranding capital, the strategies based on purchased hardware and software are considered a lock-in risk. Even when the user buys a device to serve as a public-cloud edge partner, they see the device as being linked to their relationship to the cloud provider, which they don’t see as a lock-in (even when it sometimes is).
The other problem for vendors is the fact that there’s not a lot of expensive gear involved in an edge commitment. That makes a sale less productive for both salesperson and company, making an educational sell nearly impossible, and it also means there’s a low-commitment relationship established with users, which discourages partners’ participation.
What this adds up to is that public cloud providers will likely end up owning the edge in the sense of control, and that users will likely own the devices, or at least lease them for installation on their premises. Once this trend gets started, the notion of a true edge-as-a-service, meaning an edge hosted in some facility data center, will be difficult to promote.