> Surely resolvers for a TLD with 100K domains and 1MM lookups per day are cheaper to run than those with 100MM domains and 1BB lookups per day.
This on the face of it sounds reasonable, but the cost difference is likely marginal.
If you're going to provide a highly available DDoS resistant infrastructure to deal with a large scale of customers, the majority of the costs are going to come from building that distributed infrastructure across many parts of the world.
The costs may go up between 1 million lookups vs 1 billion lookups, but if there is an increase, the order of increase will be closer to 10x than it is 1000x.
Can confirm. Our TLDs scaled by about two orders of magnitude in total registration number following our launch of .app, but the associated costs that rose were very marginal. Most of the cost is not strictly related to volume.
The increase in cost is sufficiently sub-linear that the cost of each new .org registration should be going down as the number of them goes up, if it was truly run for public benefit while breaking even.
As another comment says, most of the cost will be in building and maintaining a robust, scalable, attack-resistant distributed system. The linear scaling costs (such as bandwidth and processors) will be small in comparison.
(Of course at much higher numbers, mathematically O(N) starts to dominate again, but 1 billion lookups per day isn't enough for that.)
