Routers, switches, and other appliances are under commoditization pressure, and I’ve noted in past blogs. What about fiber and wireless access? What I’ve called “experience networks” are threatening the core and even the data center, but access is access, and no matter how commoditized experience-network devices are, they still need to blast photons over glass. Let’s look at the access/transport space now and see how players there might fare, and what they might do to fare better.
Right now, we really have four different access/transport technologies. First, we have mobile broadband, second FWA, third “wireline” physical access, and finally transport fiber. I think it’s virtually certain that this is going to reshape itself into fiber and wireless broadband, meaning consolidation into two spaces. In the long run, I think that everything that can’t be supported by glass will be supported by wireless broadband in some form, meaning that CATV will be phased out. I also think that FWA and mobile broadband will combine (more than they have already). These changes are driven by the forces I talked about in THIS previous blog.
In an experience network, which you’ll recall is also an interior network, traffic management is a bad word. The only good strategy is never to need it, because if you have to do a lot of traffic engineering to assure QoE you’re always going to be putting somebody in second place. The right approach is to oversupply with capacity, and that’s already happening. CDN and cloud providers do a lot of meshing and almost always provide more bandwidth than needed. As we move toward edge computing, metaverse, and so forth, we can expect to see more interior fiber meshing.
Interior fiber meshing creates a collision between the core router and the reconfigurable optical add-drop multiplexer (ROADM). Do you terminate optical trunks on a router as you mesh, or do you create more and more on-ramps to ROADMs? This is actually a very fundamental question, because what it’s really asking is how far out toward the edge “core” technology, meaning an optical-mesh strategy, might extend. With DWDM technology, you get a lot of wavelengths (lambdas) that can be dedicated to routes. If you start those routes further out toward the access edge, you create what’s essentially a dense and inexhaustible optical pool that requires only a minimum amount of operational intelligence. This reduces the number of electrical devices you need, and the value of network-specific software like operations/management tools.
What optical vendors would (or should) want is to push this DWDM-mesh approach as far and fast as possible. That would given them more and more of the experience network pie. But in order to do this, optical vendors need two things. First, they need to multiply the number of lambdas in DWDM as far as they can, without creating a massive capacity limit per lambda. Today, using the standard optical wavelength bands, you get a maximum of about 160 wavelengths per strand. That’s fine for experience-network core type applications, but it would break down as you go further toward the edge unless you had optical add-drop capability further out than would be normal today.
I think it’s clear that there will be DWDM and ROADM advances, but before they come along there’s an opportunity for router vendors to enhance their own optical credentials by adding high DWDM density interfaces to their devices, not only the big boxes but also closer-to-the-edge boxes. Obviously that’s being done now, at least for the big-box routers, and I think the router vendors realize that they need to get first-mover advantage here or the optical guys will expand out from the core to steal more of their market.
One area where this risk is greatest is the access edge. At the very edge of the access network, up to at least the metro point, all the traffic from the user goes inward, and there are relatively few options for routing in the other direction. Thus, these devices are simple and it’s not difficult to see how you could use optical elements. At the very edge, where actual users connect, there’s a transformation needed between core-optical and one of the edge technologies, meaning (today) PON or access fiber, mobile wireless, FWA, or CATV. If you could feed those elements with an extension of core optics, you would have little electrical-layer stuff left deeper than this point.
Just as the experience network market dynamics promote oversupply, so it happens in the access network. Customer support is expensive, and problems with quality of experience tend to fall on the access provider even if they’re actually caused elsewhere. Attracting and maintaining customers is expensive, so churn caused by QoE issues is a very bad thing to face. Over time, I think it’s clear that even access bandwidth will be boosted as much as possible to prevent issues. That doesn’t mean that access providers won’t charge more for higher capacity, at least in the near term, but I think that over time we can expect to see the range of available capacity declining, and for some technologies like FWA there may be only one “speed” option provided.
What this really means is that we’re going to see a fight-to-commoditize in the access and transport networks, a fight that likely ends up being a war-of-the-chips sort of thing. If the router vendors win, they still own the access network and electrical traffic engineering stays valuable, which means that software-level differentiation can sustain network vendor business cases. If the optical vendors win, they devalue electrical devices, devalue traffic engineering, and make the whole of the network look like that inexhaustible bandwidth pool.
The defense against optical encroachment? Metro. If we were to assume that a lot of features were hosted in metro locations, then these locations become feature on-ramps more likely to require electrical connectivity. You can’t optically add-drop through a place where features are connected, after all. The new feature focus promotes a refresh of technology in each metro location, and there are probably five thousand or more such locations worldwide. If there is an early refresh, it happens with switch/router technology and the metro becomes an electrical-layer fortress.
For network operators who are already being faced with experience-network displacement from the service space, and constrained to the access network, it means that cost management would likely become the only long-term answer. That will advance the notion of access-network federation and neutral infrastructure deployments. It will also likely advance municipal broadband, utility broadband, and “subdivision broadband”, something that cable companies in the US have been promoting with deals with developers for more than a decade.
Does this create what China is said to want, which is a kind of national super-network-and-computer utility? Maybe. The network as a capacity pool certainly could facilitate the creation of new experience networks, justified by new experiences, and there’s surely the possibility that some form of utility computing federation could arise, particularly if universally low latency created a real, broad-based, functional metaverse concept. In fact, this could well be the development that shapes the future for all of us.
For vendors, there’s a lot of transformation in the future. How far in the future? I think the concepts I’ve outlined here will become highly visible within five years, and dominant within ten years. There’s time to get ready, and to prosper in the market to come, but this is not a good time to be sticking your head in the sand, no matter where you play in the market.
