Earlier this month, I blogged about the challenges of Open RAN, but obviously RANs aren’t the only part of a telco network. Operators have accepted open-model networks and multi-vendor networks more easily outside the RAN, and so I wanted to talk about what I’ve found regarding both approaches when you leave the radio network and move toward the core.
One interesting truth is that for the last 30 years, operators have been pretty clear in saying that they tend not to want to mix sources of equipment within specific sections of their infrastructure. One said in 2002 that if they had an important interface, they were more comfortable if it connected two boxes from the same vendor. In 2023, I still find some resistance to introducing a different vendor into a network, though not quite as much and with one important exception.
The reason why operators say they’re a bit less worried about having interfaces pass between vendors is that they’ve started thinking more in terms of “mission procurements” versus “topology procurements”. The latter encouraged them to avoid introducing new vendors into an important piece of the network. The former encourages them to assign a vendor to a network mission, accepting that the mission might mean connecting to devices another vendor provided.
Some operator, notably AT&T, separated their infrastructure procurements into “zones” and wanted a series of competitors in each. This created some initial pressure on smaller vendors, but in fact there was already a trend for operators to focus on larger players. Over time, zone procurement faded a bit, and it never gained universal acceptance. Instead, we saw the start (though unrecognized) of mission procurement.
According to operators, one big reason for the shifting procurement view is wireless. As wireless networks grew in importance from a revenue perspective, infrastructure focus on them logically increased. Wireless is based on standards, but the radio network portion is generally seen as a specialty, and only a small number of vendors had the credibility to win deals. But wireless networks, beyond the RAN, were not only linked with traditional infrastructure, they often contained aggregation components that were really simply IP routers.
AT&T shook things up by announcing a commitment to open-model networking from a point close to the edge in the aggregation network, through the core. Some at AT&T told me that they believed that white-box technology, and in particular white boxes supporting Broadcom chips, had advanced to the point where the cost and competition benefits generated offset any concerns about integration or dealing with non-traditional suppliers. AT&T’s strategy, which is based on white-box hardware and open-source software, underpins the DriveNets Network Cloud. DriveNets has a number of wins beyond AT&T, and they’re the vendor I hear about most often from operators looking for an open-model strategy for infrastructure.
Open-model infrastructure, even leaving out the RAN, still faces barriers. Mission procurement demands a mission that has enough budget power to not only introduce new technology with enough mass to make a difference in the network overall, but also to fund any premature displacement of current devices. Operators tell me that this is their largest barrier to giving open-model solutions the nod. Second on the list is the reaction of the incumbent vendors, which is to discount and deal. A bit over half of the operators who said they’d looked at open-model devices, and about the same percentage of those who actually did an RFI/RFP, indicated that their incumbent vendor(s) responded with deals that made senior management less interested in the open approach.
What drives operators to overcome these barriers is one of two things, usually in combination. One is a need to upgrade infrastructure, owing to the combination of new features or capacity needs and obsolescence of existing equipment. The other is the ongoing operator challenge with profits, which is usually answered at least in part by seeking to lower costs. Almost all of the operators who commented on the latter said that they had determined that simply pitting current vendors against each other could no longer wring out enough discounts to really transform costs. Open-model networks were seen as the only option on the capex side.
Infrastructure upgrades enter the picture by lowering the cost associated with writing down existing gear to bring in something cheaper. Where no pressure to upgrade exists (and more importantly, no budget), either cost savings must be substantial or current gear has to be largely depreciated. 5G was budgeted and has generated some opportunity for upgrades (my AT&T contacts say it was the primary driver behind AT&T’s decision to move to open-model infrastructure). The problem is that, to quote one operator “5G is only skin deep”, meaning that the majority of 5G’s impact comes in the RAN or in control-plane elements and not in network infrastructure itself.
The big question, according to at least the operator planners, is whether some feature transformation of services would come along and justify a major change in infrastructure. At one point, “carrier cloud” was seen as a driver of change, but today the majority of planners say that carrier cloud is a response to a feature shift, not a driver in itself. They see IoT or the metaverse as being the potential drivers, but admit that there are shortcomings to both in terms of their realistic impact, particularly in the near term.
So far, operators have gained from IoT only in verticals that have widely distributed assets that require monitoring. Energy and other public utility applications and transportation represent the majority of the IoT applications that actually offer operators a significant new service opportunity. While some operator planners hope that applications will develop in other verticals, most admit that if that happens, it won’t be in the near term.
The metaverse situation is even more murky. Even among the operator planners there’s a fuzzy notion of what “the metaverse” is, much less what specific service technology or infrastructure impact might be associated with it. That may not be surprising given that what I’ve called the “metaverse of things” or MoT is actually related to IoT, and the requirements for what might be called the social metaverse are yet to be identified.
The conflation of these two applications and other even fuzzier applications is, for operator planners, the concept of “edge computing”. The impact of this on open-model networking, without considering a specific mission target, is difficult to assess. However, I think it’s clear that the further “out” toward the user we consider the edge to move, the lower the economy of scale advantages that could be offered and the more difficult it would be to employ the resources for applications with a wider distribution.
I’ve personally seen some utility IoT applications deployed, and those have required sensors spread over a distance of around 200 square miles, which is beyond a single tower service area. It is, however, comfortably within the scope of a “metro” area (which is roughly 400 square miles). While wireless-edge hosting would fall within RAN infrastructure’s domain, metro hosting would be a deeper-network mission. If metro hosting could be justified by opportunities in IoT/MoT and social metaverse, then this could bring data center switching to those sites, and switching is an easier mission to support with open-model network technology. The use of open-model switches would, in the view of almost three-quarters of operators, tend to stimulate the use of open routing technologies in the same area.
Edge computing, and in particular a metro-edge architecture, would stimulate the use of open devices in networks. The move would create a distributed set of hosting points that would also logically become connection hubs, and optical meshing of these points could then generate a low-latency core. All of this would create a technology refresh that could then facilitate the introduction of new, and open-model, gear. The revenues from edge computing could fund the refresh, and perhaps even fund a write-down of some legacy gear. Almost all operators agree with this statement, but they’re admittedly uncertain about whether there are business drivers for, and management acceptance of, such a move.
Even without these sorts of facilitators, I think that we’re inevitably going to see more and more open technology outside the RAN. Operators have been RAN-focused because of 5G, and they’ve been less focused on open-model networking because they’re more comfortable with traditional RAN suppliers. Now that 5G deployment is maturing, we’ll likely see a refocusing on the wireline pieces of the network, and a renewed commitment to open-model deployments.
