Aggregating rural customers further using optical technologies

[Martin Millnert]

The problem: Real rural areas have so few customers per meter/area,
that it will be a long time before it is economically feasible to
connect them for any private company.

The end-goal: Make it economically feasible to connect these networks
to commercial ISPs.

Limitations: Redundant last-mile connections to end-users is out-of-
scope

Preferences:
1) A communications operator operate a physical optical network, and
terminate the light of customers directly to whatever ISP have
equipment in this network. The communications operator *does not*
operate a layer 2, or above, network -- this is done heavily in Sweden
today and it is *not* good! (Unlimited over-subscription!)
2) The access fiber network should have diverse fiber layout from the
first cross connect node.


I am somewhat involved in a public debate in Sweden between the
government and interested networkers on how the government in Sweden
can achieve its goal of at least 90% of all households being connected
with a 100 Mbps broadband by the year 2020 (part of the IT strategy
the government has).

As a result of this, an idea came to me regarding an optimization
option for rural areas where the density even within 10km is too low
for ISPs to place equipment. Note: I've not really tried to sketch up
this network and looked at its costs yet, I'm just going by my gut
feeling at this point.

Basic idea:

In fiber to the community nets today, users usually terminate their
fiber connection well within 10km in a fiber cross connect node of
some sort. In order to increase the coverage of the physical access
network, it is possible to imagine that a number of these fiber nodes
could be connected on a ring network with high-pair-count fiber
leading to a major interconnection point, keeping distances under
~80km for all end-users, to allow cheaper long-distance gigabit
transceivers (of *today*) to be used, for example. Should optical
transmission technology evolve in the future, it is possible to
aggregate a few of these 80km-aggregating nodes as well.

Keeping one fiber pair per household all the way to the major hub
would allow for wavelength sharing to the same customer. Furthermore,
if MEMS-switching is developed further, it would be useful to stick
these into both the first fiber node, for ring level redundancy, and,
in the major hub, for automatic provisioning of service between the
fiber access network to various ISPs. Eg, each ISP could pre-provision
a bunch of transceivers in their equipment and connect it to the
automatic optical cross-connect, in order to get service and minimize
involved human labor.

My gut feeling here is that long term, and in sufficiently large
scale / cheap optical switching technology, networks such as these
would provide customers with good ISP level competition (the ISP level
competition in the open layer 2 city networks today in Sweden is very
well-functioning, symmetric 100 Mbps is roughly $15 monthly in these
networks, but the access network costs ~$35, monthly), zero over-
subscription in the access network (open layer 2 city networks in
Sweden today have scary over-subscription ratios) between customer and
ISP.


I have developed this thought further and there are a few obvious
issues that arises quite early on, thinking about this. But in order
to keep down the "TL;DR"-level I withhold this for now. (Main problem
seems to be extra costs (both CAPEX and OPEX) of running hundreds of
fibers longer distances, rather than shorter distances, and WDM comes
to mind.)


It'd be interesting to try and do a "business case" / economical
calculation on what I describe.

But before I spend time trying to do that, I want to know what people
think of this idea... and what better group to ask than the Fiber For
Communities group?

I'd be very interested in any kind of feedback.

Fiber to the people.


Thanks,
Best Regards,
Martin Millnert