The correct answer in that scenario is C should be paying for it, as in the stated scenario C’s traffic would be exceeding the peering arrangement with B and/or A, but there were/are a number of reasons that breaks down in the real world.
Peering isn’t Sender Pays, Peering is “I’ll carry your traffic if you’ll carry mine”, with the understanding that when there’s an imbalance in one direction or another that an exchange of some sort is had, be it dollars, bandwidth limits, or similar. In this case, where C interconnects with A which interconnects with B, if C’s traffic is so substantial that it’s saturating the crosslink between A and B, A would need to evaluate whether their peering agreement with C means that C needs to be paying for the network upgrade, or if there’s enough traffic moving from A’s network into C’s to offset that, and that the interconnect between A and B is the root issue. In your example, rather than paying more into ISPs and, essentially, indirectly funding US network backbone infrastructure upgrades across the board, they solved their problem with cache servers that they handed out like candy to avoid their costs to C sky rocketing. G solved this problem by buying a bunch of dark fiber which was laid on spec by contractors and started peering directly with the Tier 1 providers, dramatically reducing their cost delta.
Where Korea’s system differs is that in traditional Tier 1 peering, as I understand it, T’s ISP (call them P) should be using some of the money they get from T to pay Q and R for the excess traffic of their customer, but instead Q and R were, per the government, allowed to also charge T for delivery of their packets, resulting in T having to pay both on the up and downlink side, charging them twice for the same bit. T, rather than attempt what G did, told Korea to pound sand and exited the market.
The correct answer in that scenario is C should be paying for it, as in the stated scenario C’s traffic would be exceeding the peering arrangement with B and/or A, but there were/are a number of reasons that breaks down in the real world.
Sender pays is what Korea has. It does not work well at all.
Peering isn’t Sender Pays, Peering is “I’ll carry your traffic if you’ll carry mine”, with the understanding that when there’s an imbalance in one direction or another that an exchange of some sort is had, be it dollars, bandwidth limits, or similar. In this case, where C interconnects with A which interconnects with B, if C’s traffic is so substantial that it’s saturating the crosslink between A and B, A would need to evaluate whether their peering agreement with C means that C needs to be paying for the network upgrade, or if there’s enough traffic moving from A’s network into C’s to offset that, and that the interconnect between A and B is the root issue. In your example, rather than paying more into ISPs and, essentially, indirectly funding US network backbone infrastructure upgrades across the board, they solved their problem with cache servers that they handed out like candy to avoid their costs to C sky rocketing. G solved this problem by buying a bunch of dark fiber which was laid on spec by contractors and started peering directly with the Tier 1 providers, dramatically reducing their cost delta.
Where Korea’s system differs is that in traditional Tier 1 peering, as I understand it, T’s ISP (call them P) should be using some of the money they get from T to pay Q and R for the excess traffic of their customer, but instead Q and R were, per the government, allowed to also charge T for delivery of their packets, resulting in T having to pay both on the up and downlink side, charging them twice for the same bit. T, rather than attempt what G did, told Korea to pound sand and exited the market.