Distribute ALL Networks

Reader and  telecom expert Ted Weitz commented yesterday that terrorist threats are one more good reason for making sure that the Internet does not get increasingly centralized like the phone network.  The Internet was designed distributed so that it could survive a nuclear attack; it has no head to cut off.  That survivability has been tested in many events since ranging from the 9/11 attacks to earthquakes to more benign happening like Victoria Secrets’ spectaculars.

Ted’s logic actually applies to many more networks than just the Internet as he also reminded me in an email.  Because of  various economies of scale and power laws, networks seem to have a builtin tendency to centralization even when their basic architecture is distributed.

Ironically, as Ted points out, some of the implementation for higher bandwidth Internet backbone has meant a less distributed network with mega-switching centers whose loss would mean at least a major disruption if not regional meltdown.  On the other hand, some major Internet service providers like Google are careful to decentralize their server power so as not to create huge single points of failure even as they aggregate increasing shares of Internet traffic to their services.

Concentration doesn’t have to be physical.  If many computers are running one version of one operating system, there is certainly a danger that a virus will spread quickly and take them all down at once.  If all the routers in the world ran the same software (and many of them do), there is a possibility that they could all be attacked at the same time and in the same way or subverted so that they become accomplices to an attack.  When Google does an update of the software which runs in its distributed centers, it has to be very careful that isn’t distributing a fatal bug everywhere at the same time (don’t know how they do this but am reasonably comfortable that they understand the principal).

There is a story, possibly apocryphal,  that four railroad magnates once sat down to discuss whether the new invention, the truck, was going to be any threat.  “No,” they decided, “one day soon four of these things are gonna reach the same crossroad at the same time from four directions.  Since, there is no central control, they’ll all sit there in gridlock for ever.”  Relieved by their logic, they ordered another round of drinks.

Transportation in the US is decentralized to a fault.  Our dependence on cars and trucks isn’t very energy efficient and produces lots of CO2; BUT it is much more decentralized than a rail system.  Americans chose cars over public transport because we want to be able to go where we want when we wanted and we can afford to make that choice.  The system does give us enormous flexibility.

Trouble now is that there is a quickly increasing number of  people in the world who can afford a decentralized transportation solution and want one, too.  The network which delivers fuel for the decentralized transportation system is not decentralized at all.  Oil wells are concentrated in unstable places; sea transport is greatly focused on just a few vulnerable routes; refineries are relatively few, huge, and vulnerable to both natural and man-made disasters.  By contrast, fuel for horses almost always had a local origin and a short supply chain.

However, the same technology which allows us to run a decentralized communication network like the Internet and which can allow us to move to a mesh architecture for Internet access, also can help with decentralizing our energy supply.  Continued here.

Previously blogged about a decentralized Internet as America’s Antiterrorism Network (here and here).

The post about Internet Centralization which Ted commented on is here.