Why Is Carrier Cloud on the Back Burner for Carriers?

I noted in my blog yesterday that I was surprised and disappointed by the fact that the network operators didn’t seem to have given much thought to carrier cloud in their fall technology planning cycle.  I had a decent number of emails from operators, including some that I’d surveyed, to explain why that seemed to be (and for most, was) the case.  I thought I’d share a summary with you.

The net of the comments was that for better or worse, the operators have come to view “carrier cloud” as the outcome of things they’re doing rather than as a technical objective.  About half those who commented to me said that they believed that over time as much as a quarter of all their capex would be spent on servers and cloud technology.  However, they were all over the map in terms of how they believed they’d get to that point.

NFV remains the favored technology to drive carrier cloud, despite the fact that there is relatively little current indication that the connection exists, much less is growing.  This is in sharp contrast to my own modeling of the carrier cloud opportunity, which says that nothing will happen with NFV/carrier cloud in 2017 and that only about 4% of carrier cloud opportunity in 2018 comes from NFV.  In fact, the attitudes on carrier cloud demonstrate how difficult it is to survey users on many technology trends.  Two years ago, an even larger percentage of operators told me that in 2017 NFV would be driving carrier cloud.  My model always said “No!”

The second-favored technology to drive carrier cloud is 5G, and interestingly the percentage of operators who say 5G will be the driver in 2020 is almost exactly the same who said that NFV would be two years ago.  The majority of this group still think that NFV is the real driver, and they believe carrier cloud comes about because of NFV involvement in 5G implementation.

It’s really difficult to say what 5G would do for carrier cloud, largely because it’s difficult to say what 5G will do overall, both functionally and in a business sense.  A third of the comments I got from operators that favored 5G as a carrier cloud driver admitted that 5G “has a long way to go” before real adoption can be expected.  In other dialogs I’ve had with operators, they indicated that their current 5G planning focused on the radio network (RAN).  Some said they wanted to extend FTTN with 5G instead of DSL/copper, but most thought they’d do the upgrade for competitive and capacity reasons.

Those who think 5G boosts NFV, which boosts carrier cloud, are thinking mostly of a 5G goal of making network service features interconnectable and “roamable” to the same extent that connectivity is.  The problems with this vision are 1) there is no currently approved approach for VNF federation in NFV, 2) there’s no significant VNF deployment except in premises devices, 3) many operators don’t like the notion of constructing services from components like that, fearing it would eliminate a large-provider advantage, and 4) we still don’t have a 5G standard in this area (and probably won’t get one till next year).

The actual place where 5G might help carrier cloud is in the residential broadband space.  I’ve been blogging for almost a year on the fact that operators told me the most interesting 5G application was DSL replacement in FTTN deployments, and Verizon has now announced it will be starting to deploy that way halfway through 2018.  Clearly the thing that needs 5G capacity versus DSL capacity would be video, and video turns out to be the thing my model says is the best near-term driver of carrier cloud.

In 2017, video delivery enhancements and advertising caching (CDN and related tools) accounted for almost 60% of the opportunity driver for carrier cloud, and you’ve got to go out to 2020 before it drops below the 50% level.  Obviously there hasn’t been much uptick in the adoption of carrier cloud for video/ad hosting, but here’s an important point—you can’t deliver traditional FiOS video over 5G/FTTN.  You’d have to stream; thus, it is very likely that the Verizon-style 5G/FTTN model would require augmented caching for video and ad delivery.

The good thing about this particular carrier cloud driver is that it would very likely create a demand for edge-caching, meaning edge-hosting, meaning edge-concentrated carrier cloud.  FTTN terminates in central offices where there’s real estate to establish carrier cloud data centers.  These data centers could then be expected to serve as hosting points for other carrier cloud applications that are not yet capable of justifying one-off deployments of their own.

By 2020, when video/ad support finally drops below 50%, the biggest uptick in carrier cloud driver contribution comes from the 5G/IMS/EPC area, meaning the virtual hosting of 5G-and-mobile-related elements.  This is because as wireline DSL/FTTN is replaced by 5G/FTTN, there’s certain to be symbiotic use of that home 5G.  One of the easiest ways to stall out a new RAN technology is to have no handsets capable of using it, which happens in part because there’s no 5G cells to use those handsets in.  If many homes have local 5G, then why not let those same 5G connections support the homeowner?  In fact, why not let those FTTN links to 5G-for-home also serve as 5G RAN cells for mobile services?  You end up with a lot of 5G deployment, enough to justify handset support for 5G.

The big carrier cloud opportunity starts to show at about this same point (2020) and by 2022 it makes up half of the total carrier cloud driver opportunity.  It’s the shift to event/contextual services, including IoT.  The edge data centers that are driven by 5G/FTTN are available for event processing and the creation of contextual, event-driven, services that most cloud providers won’t be able to supply for lack of edge data centers.  This is what finally gives the network operators a real edge in cloud services.

Of course, they may not take the opportunity and run with it.  You can fairly say that the big problem with carrier cloud is that it’s driven by a bunch of interdependent things and not one single thing, and that’s probably why operators don’t think of it as a specific technology planning priority.  They need to learn to think a different way, and I’m trying now to find out if there are real signs that’s going to happen.  Stay tuned!