Revisiting the PON vs P2P debate
A year and a day.
That's how long ago I first posted about PON and P2P (The Point of PON). Unknowingly, with my first "content post", I'd stumbled upon the great debate of the fiber to the home space. My newbie comments on this heated issue got me picked up by a number of bloggers, either for or against which in turn pushed me to write two more posts on the topic (More on PON and Let there be a debate!)
A year and a day onwards, what's changed? At face value, not much. The two technologies are still being bandied about like the Holy Gourd and the Sacred Sandal, and most of what I read or hear about comparing the two amount to a lot of faith and only a tiny little rationality. This is still the most heated debate in the FTTH space. I have learned a few things though that I think are quite interesting to share:
- Don't confuse technology and topology: One of the great confusions of the PON/P2P debates is that PON is a shared infrastructure and P2P is a dedicated infrastructure. While that's true on the active level, it doesn't have to be on the passive level. In other words, we should distinguish passive star/homerun and PON/P2P. You can have one fiber running from the CO to every home and still implement PON. In fact, that solution has distinct advantages of scalability in a PON architecture: with curb-side PON, you need to provision a full equipment to host 32 or 64 customers before you've signed the first customer, and may have to keep that in place even for a single customer. If you PON equipment is in the CO, you can scale up as you sign customers. You can even use Ethernet P2P for selected customers if needed (like businesses...) But that requires a homerun infrastructure.
- Homerun vs Passive star: So now there's two debates, not just one. And the passive infrastructure debate, in my mind, is mostly clear. Whenever possible, you should deploy homerun. It's absolutely future proof, it allows for the best flexibility and innovation in terms of services should you wish to go down that route, and it's the least likely solution to require twists and loops in your infrastructure should a nosy regulator suddenly ask you to share your infrastructure. It's 10% more expensive, but it's a true 40-year investment. Now what do I mean by "whenever possible". In greenfield deployments, go for homerun, no questions asked. In brownfield deployments, if you absolutely need to deploy fast, and there's no available duct or pole space, you may have to go for a passive star. Certainly municipal projects, since they usually have easier access to ducts and poles and furthermore have greater opportunity to dig, should always favour homerun.
- Cost differentials should be looked at at the active layer: Unless you're a dark fiber ftth provider (in which case I want to hear from you!), looking at the difference in passive costs doesn't make sense. While a passive star is undoubtedly slightly cheaper to deploy, the splitters from a PON solution degrade the signal and require more powerful lasers on both ends. In other words, when you factor in the CPE costs, you will find - roughly - that on the CAPEX side, passive star+PON=homerun+P2P. That's assuming you don't get clever in the ways suggested above!
- OPEX savings: That's still a big bag of unspecified hype to me. I've heard people saying there was less OPEX with PON, others saying there was less with P2P, and I have yet to see any definitive study from a neutral party that demonstrates a difference in OPEX costs between PON and P2P. Stress on neutral. There's no doubt (and demonstrated evidence) that there are significant OPEX savings between an aging copper and a brand new fiber network, but beyond that... I don't know.
- PON and unbundling: That was the core of my argument a year ago: incumbents choose PON because, amongst other things, it's hard, and maybe impossible, to unbundle. If I was to rewrite that today, I wouldn't be so bold. A more accurate statement would be that deep-splitted PON is definitely complicated to unbundle, and can even designed to make unbundling impossible. PON vendors say that the same can be said of P2P, which is true in theory. In practice, a homerun infrastructure makes unbundling a lot more manageable provided the COs are dimensioned correctly to do so. Regardless, regulators in general and Vivian Reding in particular are examining at PON with a none-too-happy gleam in their eyes for that very reason, and whether right or wrong, it's something that will need to be taken into account, especially if regulation goes beyond frowning and actually regulates.
