|
Doc Searls has an essay about bringing fiber optics to every home in America. It is aimed in the right direction, but makes a couple of mistakes on the numbers and falls to ground way short of its target. It troubles me that I appear to be the sole source for Doc’s numbers (on the basis of some informal conversation and my Telecom Day speech in Wellington NZ last May).
This post is an attempt to correct the record, and to create one where my previous thinking has been private.
Doc writes:
A typical fiber trunk fits an 864-fiber cable inside a 1.5-inch conduit. Each fiber can carry 10 gigabits of data. The total comes to 1.6 terabits. Here’s how David Isenberg puts that into perspective:
If all 6.5 billion people on earth had a telephone, and if they were all off-hook, generating 64 kilobits a second, and all those conversations were routed to this cable, there would be 100 fibers still dark.
Almost. Each fiber, in fact, can carry 10 gigabits per second **on each of 160 wavelengths** using today’s off-the-shelf technology. In other words, each fiber, lit like this, carries 1.6 terabits. The arithmetic shows that 10 gigabits times 864 fibers is 8.64 terabits—wrong. The actual computed capacity of an 864 fiber cable, lit as described above, is 864 times 1.6 terabits, or 1.4 petabits.
It is very bad to trust me on simple arithmetic; I have been known to think about dividing by six while I divided by four. I have been known to divide when I should be multiplying. And I’ve made plenty of oversight mistakes. But I’ve worked this particular problem a few times from scratch, but please check my work…
I’ll leave it to the reader to calculate whether 6.5 billion people on earth, each generating 64 kbit/s at the same time, with all that traffic going to a fiber cable in which each fiber is carrying 10 gigabits per second on each of 160 wavelengths, whether all that traffic could squeeze into 764 fibers.
Another quibble. Doc writes that a *typical* fiber cable is 864 fibers. Typical? The fiber in front of my house is a 144 fiber Sumitomo cable. I saw it installed. My understanding is that 864 fiber cables are mostly used for long-haul these days. The example of 10 gigabits by 160 lambdas by 864 fibers is only *typical* of what’s possible. You can light more than 160 wavelengths, and jam more than 10 gigabits down each one. I saw a Sumitomo fiber cable with 2000 fibers in it in 2004. You can have more **or less** of any of these quantities.
Here’s another item from Doc that needs a bit of post-hoc sharpening: Doc writes,
Bringing fiber to homes and offices costs between $1000 to $7000 per “drop.”
Yes and no. I’ve been using $2000 per urban/suburban home and $6000 per rural home as a rule of thumb, a first approximation average cost. This includes homes hooked up for 50% of the homes passed. It is an impressionistic amalgam of lots of stuff I’ve been told in private conversation with fiber mavens, and it has been verified as reasonable by several people who have actually built fiber networks. [By the way, others, telco types and Washington DC policy people, tend to cite lower numbers, closer to $1000 per home, but when I dig, this is only the cost of passing the home, not hooking it up.] But it is also worth noting that the costs can be very much higher than $6000 in individual cases, urban or rural.
So, if you assume 100 million urban and dense-suburban residences in the United States at $2000, and 15 million rural homes at $6000, that’s $290 billion. (Heck, the U.S. found $700 Billion in a few days last month cause we had to have it now… but we still don’t know what we’re going to use it for.)
Doc didn’t get anything overtly wrong when he wrote:
David Isenberg tells me a good sum to invest is $300 billion. That would be for “every home passed with more density than about four per road mile and a 50% take rate.”
But he sure ignored a lot of relevant detail.
I’m nitpicking, right? Well, how do we know whether Doc missed another factor of 160 (or its conceptual equivalent) when he concludes (perhaps based on another unvetted morsel from another sole source) that a public-private partnership can build all the bandwith for any purpose its users want to use it for, even as it
Sorry, Doc. We can’t pretend there’s no incumbent, no installed base or that the market is always right. The intellectual heavy lifting in the Fiber to Every Citizen effort will be in getting the details right. This includes how the new infrastructure is regulated. Yes, regulated with laws that say, “No snooping, no blocking, no discriminatory pricing or treatment, and what goes in is what comes out.” We don’t need any more Enrons, Blackwaters or AIGs. We need the private companies to partner with the public, not the reverse. We need to get the incentives right. We might even consider whether public-private-partnership is the right model at all. But, in any case, we need to do the math.
Sponsored byWhoisXML API
Sponsored byVerisign
Sponsored byDNIB.com
Sponsored byRadix
Sponsored byVerisign
Sponsored byCSC
Sponsored byIPv4.Global
I seem to recall reading that Verizon had reduced their cabling costs to around $700 per home passed, presumably in typical suburban neighborhoods, but I’m not sure about that.
There are around 120 million homes in America, about 75 million of which are detached, single dwelling units (regular houses, not duplexes, condos, or apartment.) About 25 million are rural, so that leaves about 20 million apartments, etc. I figure we can assume that the rural homes are more expensive to wire and the high-density housing is cheaper to wire, but I don’t know if these two factors balance out. If they did, stringing the wire would run you $84 billion. But that’s dark fiber, so you gotta light it up, and that’s where speed makes a difference.
You can literally pay as much as you want to pay for the gear that lights up this fiber. And who decides how much to pay?
Dave Burstein did some alternate calculations about the more general question of bringing 100 Mb/s connectivity to each home. The cheapest way to do that, obviously, is on the existing coax cables. Now you only have to pay for the networking gear, and you’ve saved the $84 billion. So that also is a question of your timeline and planning horizon.
How much do you want to pay today for the connectivity of tomorrow? Presumably, you can get the Treasury to pony up as much money as you want if you simply couch it as a telecom industry bailout.
The fiber I’ve installed was all simplex, so two fibers per circuit (Tx/Rx). Is that also the case here? I know that the technology exists to run full-duplex on a single fiber, but I have not seen it in common use (perhaps for undersea fiber?).
In this case then, 864/2 = 432 connections; at 160 lambdas that’s 69,120 circuits, each of 10Gb is 690,120Gb or 675 Tb. (I think I did the math correctly…)
I realize this is for purposes of illustration/amusement/amazement, but I think it unlikely that more than one wavelength would be used per fiber for a residential drop given the added expense of the “prisms” for selecting the correct wavelength. (And then, how could you be sure that someone else wasn’t listening in on *your* frequency…?)
Anyway, all very fascinating.