Regulating Network Peering
The economic forces behind the internet remain, for me, the largest mystery of what makes the whole thing go. I understand the client/server relationship, posses a decent grasp of the seven layers of network communication, and even get DNS. But how companies who actually own the network make money is something that has always puzzled me.
As part of the server collocation service I use for LegSim I pay a fair chunk of the monthly bill for bandwidth. I assume the facility then takes that money, skims off a percentage, and uses the rest to buy bandwidth from someone else… who in turn does the same with all of the various networks it connects to, and so and and so forth until all of the money is dispersed to all of the networks on which my users reside.
The only problem with this scheme is transaction costs (as with so many things in life). Monitoring the connectivity of every network to every other network must be a massive undertaking, resulting in tremendously complex billing. Which is why networks often use a billing method called peering. With peering, networks of similar sizes agree to charge eachother nothing on the assumption that they are both using eachothers resources equally and thus it will all cancel itself out in the wash–peering does it all without the transaction costs.
Seems this model is beginning to breakdown. Last week two sizable network owners disagreed on whether they were comparably sized. As a result, one of them shut off connectivity to the other until the other forked over payment. That didn’t end well for either of them and now my telcom hero Congressman Rich Boucher of Virginia is telling the industry he plans to introduce legislation to allow the FCC to regulate peering deals. Supposedly the new FCC power will be limited to traffic cop status, allowing them to resolve peering disputes between network competitors. But if the history of the FCC is any indication, I doubt it will take very long for the traffic cops to start erecting checkpoints and jersey barriers in the name of network reliability.