The point of PON

Here in France, there’s an expert debate on the issues related to the technological choices of deployment architecture. I’m not talking about the relative merits of FTTH versus FTTB (or FTTC), that will be a topic for another post, most likely. I’m talking about the choice of PON versus point-to-point architectures. The topics are obviously related, and I'm not certain I understand all there is to understand about these issues (by far), but the purpose of this post is to lay down what I do understand, and hopefully the gods of Web 2.0 will shine on me and contributors will fill in the gaps.

Just to clarify my purpose, I'm not interested in exploring the technological choices per se, I'm only interested in understanding how it may affect the customer and/or the business model of the operator making that choice. In order to make this as clear as possible, I'll start by describing, as best I can, the technologies involved and their impacts.

  • The idea of PON (Passive Optical Network) is that the last active element of the network is connected to the backbone. Beyond that point, a passive optical splitter distributes the bandwidth amongs connected homes. PON is a choice ideally suited to broadcast (because you're sending the same information to each customer), and historically most cable operators used PON or PON-like models to deploy their TV services. The main advantage of PON is that it requires less wiring than point to point, that it mutualises the service for several subscribers, and that it has no active element beyond the central exchange, hence more stability. The main drawback of PON is that it mutualises the bandwidth, and that therefore the uses of one subscriber may affect the experience of other subscribers on the same branch. This is a well known problem of data services over cable where heavy P2P usage by a few hog the available bandwidth for the many.
  • At the other opposite of the spectrum is Point to Point (also called Direct Fiber). This model links each subscriber home directly to the Central Exchange through an Ethernet connection - a model that is not unlike the existing copper infrastructure. The main advantage of that is a dedicated line from subscriber to backbone, hence the capacity to provide optimal service. Its main drawback is cost, since it involves more important lengths of fiber and more ducts as well.
  • In between these two models are various means of splitting the signal between various customers through Active Ethernet equipments. These models allow a level of mutualisation and the associated costs savings, but introduce active elements between the Central Exchange and the subscriber, hence higher equipment and maintenance costs and the risk of less stability.

Ignoring for a moment the grey area in the middle that this last option provides, it seems that the choice between PON and point-to-point is a choice between a relatively guaranteed QoS and optimal customer experience (at higher costs) and a less costly deployment (with a risk of degraded quality). Until recently, this was my understanding of the equation and I was therefore not surprised to see operators annoucing one or the other depending on their image, investment capacity and ambitions.

Last year, though, the French government purchased an in-depth study on what is called Very High Bandwidth (Très Haut Débit) here in France. The study (Etude sur le développement du Très Haut Débit en France) was undertaken by French consultancy Idate and includes an economic evaluation of the deployment costs. The model used postulates greenfield deployment and a targeted 40% of population covered by 2015, with an estimated 18% of homes subscribing to a fiber offer by then. I will most likely study the model more in-depth in the coming weeks, but one of the interesting conclusions is that the differential cost between PON and point-to-point (specifically GPON and EP2P) was much lower than anticipated. An individual  EP2P connection costs 950 EUR whereas a GPON connection costs 879 EUR.

According to this study, point-to-point is a mere 8% more expensive than PON. Not that an 8% cost reduction is to be scoffed at, but considering how some cable operators have seen their image considerably degraded by perceived lack of QoS on data services, it doesn't seem to be high enough a difference to justify going for the riskier alternative. All the more so since the price difference is lower in dense urban areas, where most players are likely to deploy first. Furthermore, point-to-point is often perceived as a more long-term solution because it's dedicated to each subscriber's home.

And yet major players like NTT or France Telecom are currently deploying PON. I asked everywhere I could to find a convincing reason for this choice, and I think I may recently have been pointed out to me. As usual with these things, it seems pretty obvious once you envisage it: one of the characteristics of PON - being a mutualised infrastructure - is that it is virtually impossible to unbundle from a technical standpoint.

Here in Europe, announced FTTx projects have often been delayed by the stance of the European Union on unbundling of fiber networks. Incumbents defend the viewpoint that the fiber playing field is level, since they have no prior infrastructure and are beginning at the same time their competitors are. The EU's stance seems to be that any player in a dominant position should open their networks, regardless of how recent said networks are.

By choosing PON, are incumbents ensuring that even if the EU finally legislates on fiber unbundling, their technical choices will render the decisions inapplicable? If I was in their seat, I'd probably do the same, and accept a little degradation of service for a increased safety in corporate strategy. But maybe my analysis is flawed, and I'm missing something entirely.

What do you think?