Slashdot Mirror
International Internet Infrastructure Triples
bda writes: "TeleGeography has just published this year's statistics for international Internet infrastructure growth, aka how much capacity goes where. Worldwide, Internet bandwidth nearly tripled (174 percent growth), but behind it are some pretty big differences -- growth ranged from 90 percent (less than doubling) for Africa to 479 percent (almost sextupling) for Latin America. City-wise, the top interregional hubs connecting between continents were New York, London, Amsterdam, Paris, SF, Tokyo, Washington DC, Miami, Los Angeles, Copenhagen, in that order. So the Internet is still fairly U.S.-centric ... but still becoming less so. Asia-Pac's ratio of out-of-region to in-region international capacity went from 7:1 to 4:1; Lat Am's from 36:1 to 7:1. The most obvious factor in long-haul Internet bandwidth growth seems to be whether or not someone has plunged ahead and laid dark fiber. When we looked at trans-Atlantic and trans-Pacific capacity, Internet capacity stayed pretty constant at 10 percent of what was theoretically possible over lit fiber." You can read the executive summary (pdf), or you can (gulp) pay $1,995 for the whole thing. That would work out to about 50 copies of the Atlas of Cyberspace.
Wouldn't it be worthwhile for the researcher to add the Inner-European Interchanges, too? I remember from my time in Germany that FrankfurtLondon is a MAJOR Hub, for instance many English Providers have a direct peering at the DECIX. Same goes for the Netherlands and other countries. Now it depends on your routing, but many times you are being routed through other European countries first and then it very well matters how your interconnect with those countries works ...
If x increases by y percent, then
new x = (old x) * (1.y)
In other words, you count the 100% that x originally was, so sextupling (~579%) is correct.
Jaime
There's lots of glass in the ground now, most of it dark.
Digging trenches and laying conduit is only part of the story. It is expensive to do, and costs enough that if you're digging in the first place it's not a whole lot more expensive to lay a hundred fibres as it is to lay one. It's cheaper to overprovision than it is to go back and dig up your trench in three years.
Laying the fibre is only part of the question. Getting the fibre into the ground costs (grasping at an arbitrary number) a quarter of the total cost to light it. All that fancy kit on either end and points in between costs lots of money and sucks a lot of power (not an issue in LA or NYC, but a major issue if you have to regenerate your signals somewhere in the middle of the Rocky Mountains).
Nortel and other systems providers will eventually start to make a decent income again selling equipment to light the dark fibre (or add wavelengths to partly lit fibre). It's Corning I'm worried about...
.siggy