Everyone knows that network operators want to manage costs. The essential challenge they have is that the Internet model of networking and information/content distribution doesn’t pay a consistent rate for capacity, but Internet use, driven by new over-the-top (OTT) businesses are based on creating more demand for it. Thus, expanding use of the Internet will naturally drive down revenue per bit.
There are two components to operator cost; capex and opex. Capex has two basic sub-components—sustaining budget and project-based new spending. Keeping current networks running is the first and things like 5G upgrades examples of the second. Opex has many elements, but most of the attention goes to what I’ve called “process opex”, which is the cost of the processes that sustain proper operation of the network and operators’ overall business, which is the OSS/BSS/NMS stuff.
Operators tell me that, historically, they’ve handled the financial management of their businesses in much the same way that enterprises do, meaning that they will assign a capex and opex budget based on current services, and any increases will have to meet company ROI targets. The only differences operators contend with is the fact that some operator technology has a longer (sometimes much longer) period of expected useful life. In most cases, operators report that new projects that could justify increases in capex or opex are projects designed to raise revenues, meet regulatory requirements, accommodate changes in standards, etc. Here, there’s a difference with enterprises as well; most new enterprise projects approved target an overall reduction in costs somewhere; slightly less than a quarter of enterprise projects are justified by higher revenue levels.
If you have long-lived infrastructure, the general expectation would be that you have a correspondingly long planning cycle. That used to be the case for operators, and some projects (like accommodating a new wireless standard like 5G) will still take a long view—ten years or so, in contrast to a three-year period typical for enterprise planning. It’s relatively rare, say operators, to be looking at long-term network trends or technology trends to “optimize” networks; this sort of thing would more likely accompany another driver, again like 5G. In between these drivers, the presumption is that the framework of the last major project is to be sustained.
Today, some operators are questioning that. Of the 88 operators I chat with, 28 say that they believe that there should be a review of longer-term network evolution done at least every two years, and that this review should aim to identify cost-saving projects that likely involve some reshaping of the network/business framework, before the next major-project initiative resets thinking. Their reason is simple; their market is set by short-cycle planning of other organizations, and they need to regularly refresh their framework vision to correspond with these OTT trends.
What sort of things are those 28 operators looking at, or proposing to look at? The comment that comes up most often is simplification, followed by open. The technologies most often mentioned are optical and white box. DriveNets, whose cluster-modeled core network devices were introduced in an AT&T win and who recently won a deal with Comcast, shows up in spontaneous comments from operators for the first time, as an example of an open approach. The original core-centric mission is perhaps less a focus than something a bit broader.
DriveNets builds up network nodes from a cluster of white boxes running advanced Broadcom chips and custom software. One of the operators in my group of 28 said they thought the ideal network model would be an all-optical core, DriveNets nodes as edge-of-core, and then the same white boxes distributed outward, connected with the fastest possible fiber and, in the inner layer of the access network at least, multi-homed to the edge-of-core elements. This would give them a common pool of spare parts and a lot of capacity to play with.
The approach obviously simplifies, both by reducing the number and variety of boxes, but also reducing the features needed for capacity management. We’ve had many features and protocols that aimed at handling overloads and failures, and the more features you have in a device the more things can go wrong in setting them up. As user connection speeds rise, fiber becomes essential further out toward the edge, and operators are starting to realize that this can be leveraged to boost capacity to the point where congestion is less likely to happen at all, making all those management features redundant.
The question is how you evolve to this sort of model, whatever vendor you pick. Vendors, operators say, tend not to advocate something like this because they fear that the selling cycle would be too long and the risk of competition too high. Even the operators who mentioned a specific vendor in their comments suggested that vendor was not pushing an extreme do-over. Interestingly, the operators didn’t suggest that they wanted one immediately, only that they wanted to articulate a long-term model that they could then evolve toward.
This, to me, is interesting because it suggests that the operators trying hardest to do the right thing in the long term are still being hampered by short-term thinking, but on the part of their vendors more than their own management. Is this a problem with vendor sales management, the sales force itself, or higher company management? It’s hard to say from the operator comments, but I’d say that the comments suggest that it may be all of the above. Setting, and more importantly emphasizing, quarterly quota goals tends to focus everyone on their feet. A sales force will sell what, and how, they’re commissioned to sell. What I’m seeing, or at least seem to be seeing, is that setting quarterly goals to meet makes a sales force reluctant to look at longer-term evolutionary goals their buyers might be trying to address.
One particularly interesting point here is the 28 operators’ reaction to MWC. All of them, using one term or another, characterized the focus of the show as being either too “tactical”, meaning vendors treated it as a product bazaar, or too “theoretical” meaning it talked about things like “Open RAN” or “AI” without linking the terms to any specific evolution of infrastructure. “You can’t buy a standard,” one operator noted.
It’s really difficult to assess the extent to which vendor short-term quota and financial reporting obsessions interfere with long-term operator infrastructure planning, but I think that the comments they offer suggest that the impact is significant. The question is whether anything can overcome that influence. One candidate that’s now emerging is the use of AI by online network publications to deliver insights gleaned from their history of coverage. I’ve tried one that’s in beta, and unfortunately all it delivered was 404 errors, so I can’t say whether there’s any useful insight in there to be gleaned, or if the mechanism is workable. I don’t think there is really any media drive to support long-cycle tech planning in any case, and it may be that such a drive is the only way we can help operators balance all the forces that complicate their businesses.