First/Last Mile: Optical versus RF versus copper
jamesr at best.com
Sun Apr 27 13:06:13 PDT 2003
I spent part of Friday evening talking through some of the suitability
concerns for the many "last mile replacement" networking technologies with
the marketing VP at my networking company. One primary concern was that for
many types of markets, particularly suburban markets, viable ubiquitous
last-mile solutions haven't really been commercially realized yet. Note
that for our last-mile bandwidth models, we are looking at a minimum of
2-Mbps average to the backhaul per household with a substantially higher
After some discussion and thought, we developed a networking product concept
that appears to meet all the requirements for ubiquitous last-mile
replacement for the suburban market (low unit cost, high effective
bandwidth-to-core/person, and low installation cost). This can't possibly
be a new idea, but a Google isn't giving me much either. Since there are a
number of FoRK-ers with interests in alternative networking technologies, I
thought I would give the idea a shot here first. If I didn't miss something
obvious, we might put some engineering effort into trying it out.
The general networking technology options currently available are:
Fiber Optics: Best networking there is, but way too expensive to deploy.
The cost is largely regulatory and having to do with using public spaces to
put the fiber down in the first place. The only markets where fiber optics
are viable for ubiquitous networks are ultra-dense urban areas like
Manhattan or downtown Chicago where lateral costs *per person* would make
installation reasonable. It could be viable in some suburbs if the
governments gave free and unfettered access to public spaces.
Wireless RF: The bandwidth/acre available using wireless networking
technology (at least as is possible under current regulatory guidelines) is
too low to be viable in most places people live. However, it rocks for
rural or "edge of town" type markets that are usually under-served for
high-speed data networking anyway. It also makes a good access point
technology for very localized bandwidth distribution.
Copper: Perfectly serviceable, if the telcos didn't have a lock on this
resource. But if you were going to put in your own copper (since you can't
use telco copper for most intents and purposes), you might as well put in
Free Space Optical (FSO): Very fast (second only to fiber), but also pretty
expensive. It also has relatively limited range in models with "child-safe"
laser power, around 1000 meters (halve that if you want environment proof
links). Even cheap wireless goes much farther. Nonetheless, it makes good
point-to-point backhauls where fiber is impractical.
If you investigate these options, one thing becomes apparent: Free Space
Optical gear is *grossly* over-engineered for many possible uses. For
starters, much of it uses telco interfaces. Many of them are designed to
act as the physical transport layer for SONET-based fiber links at
OC-12/OC-24 data rates. Furthermore, most of the units are designed to work
at ranges of no less than 1-km, which requires the addition of stabilizers
(e.g. gyroscopes) to compensate for building sway and similar. Beam
displacement caused by building movement is magnified as a function of
distance. A building motion that would not move the beam off the sensor at
100 meters would cause the beam to completely miss the receiver unit at a
1,000 meters. Lastly, the use of eye-safe lasers (i.e. lasers at power
levels that won't cause permanent eye damage) limits the "burn-through" of
the link in bad environmental conditions (e.g. rain and snow) regardless of
how stable the laser is or how much beam shaping you do.
Virtually all the "expense" of decent FSO units are the engineering and
interface add-ons to make a dirt cheap diode laser a telco-class physical
layer solution with extraordinary data rates and long usable range. If you
wanted to, say, move 10-100 Mbit/s a distance of 100 meters, you don't need
any of these features or add-on capabilities.
So the idea was this, based on some basic research suggesting that it is
Build short-range eye-safe diode laser mesh (or linked ring) nodes, with a
design range of not more than 200 meters, and with either 10 or 100 Mbit
link speed and with Ethernet TP coming off the mesh or ring router. Because
the operational range is short (200 meters max, but it could be even less)
and the unit would use proper diode lasers to drive a modest data rate, you
get the following benefits:
- Trivial alignment, making installation cheap. True line of sight without
having to worry about Fresnel interference.
- Building sway and other alignment issues magnified by distance become
non-issues for the most part.
- Use of diode lasers at these ranges gives excellent environmental
burn-through, but eye-safe units at the edge of the visible spectrum or
below it eliminate safety concerns.
- High theoretical bandwidth (extremely high on a per unit area basis) and
superior latency performance. Wire-speed should be possible on the cheapest
units. If you need a faster link, channel-bond multiple lasers.
- Mesh or linked-ring routed network compensates for drop outs caused by
beam interruption. These would actually be almost ideal for linked-ring
topologies, which require more modest routing logic (read: cheaper).
- The RX/TX units would be quite small, certainly no larger then RF
subscriber units, and use less power than most RF units.
- Cheap, cheap, cheap. A mass produced node shouldn't cost more than a
couple hundred bucks.
So am I missing something obvious as to why this won't work? We've used all
the networking technologies I mentioned above, so I am familiar with the
capabilities and limitations of all of them. Even "cheap" FSO tends to
spend a lot of money and effort on effective link range, which degrades
rapidly with distance. By intentionally designing the system to only be
used for very short ranges (by FSO standards), 95% of the engineering effort
disappears, but everyone seems interested in producing systems that can
maintain extremely fast links at ranges of no less than 400 meters.
Is this a reasonable design concept, or is there a glaringly obvious problem
that I missed? That FSO systems are already in use around the world
suggests no, and that it is more of a market oversight or perhaps the
nascent state of mesh-type networks (which typically use wireless) which
hasn't really gotten around to exploring this option, particularly since the
gear doesn't really exist.
Comments? Suggestions? If this concept is as viable as it looks
(short-range laser rings/meshes), we might seriously consider building units
jamesr at best.com
More information about the FoRK