DSL Prime, DOCSIS Report, Fiber News, A Wireless Cloud

I made a significant error two years ago drawing the conclusion that GPON was much harder to unbundle than active, point to point Ethernet. I particularly apologize if Vivenne Reding was influenced by what I said at the time. 

The easiest way to unbundle GPON is "bit-stream" where the competitor connects behind the service provider's OLT. This can offer a reliable 100 meg to all customers because there's plenty of capacity in the standard 2.4/1.2 split 32 ways. Until any likely traffic model, that's 99% reliable at 200 megabits in both directions and much more so at 100 megabits. The trick here is to understand that the cost should not vary with traffic; the charge to the alternate provider should be for one connection at "up to 100 meg" rather than per gigabyte of traffic.. At a 1.5 to 1 oversubscription/contention ratio, there just isn't any congestion to worry about and hence no real cost for the higher speed/capacity. The result, if competition isn't too weak, will be higher speeds/caps for consumers at no additional local infrastructure cost. Since the backhaul/peering/transit is so inexpensive ($1/month typical), this will mean high speeds will cost only a little more than low speeds. 

ECI's paper points to two other ways to effectively unbundle. The first is for competitors large enough to support a dozen or more subscribers from GPON splitters in the field. GPON requires only a single fiber, but there are almost always plenty of spare fibers. There's no technical reason the competitor couldn't simply pay for that fiber and install their own splitter for a dozen or two customers. This is essentially "sub-loop unbundling".

ECI is also suggesting that the 2.4/1.2 bandwidth of the OLT could be split between several companies directly. For example, one carrier could buy 500 megabits down, a second 200, and the incumbent would retain 1.7 gigabits. The carrier who bought only 200 megabits might offer a 50 megabit service at a 4:1 contention ratio to 16 customers, which should work fine. The carrier with 500 meg might offer 100 meg service to 20 customers with a similar low ratio. This requires substantial work in the OLT, but ECI equipment has traditionally been very robust enough to handle demands like that.

Google's gigabit service will probably be active Ethernet, which retains many advantages. ECI's white paper is http://www.ecitele.com/Products/FiberAccess/Fiber%20Access/Open_Access_and_Local_Loop_Unbundling_on_GPON_Networks.pdf

(Note: While publications often charge for linking to white papers, ECI didn't pay for this. It was interesting enough for me to make an article. I also offer white paper distribution as part of advertising packages)