DNS Infrastructure Consolidation Study

This project revisits and extends research originally published by researchers at the University of Chicago, examining how DNS infrastructure is consolidating across the internet. I analyzed the top 10,000 domains from the Tranco ranking to understand how much control a small number of providers now have over global name resolution.

The results were striking. Amazon and Cloudflare alone now exclusively host 47.9% of domains — up from 37% in the 2022 study. The top ten providers account for 61.9% of domain reachability, a sharp increase from 50.7%. That level of concentration raises real questions around resilience, cost pressure, and long-term control of internet infrastructure.

To run the study, I built a four-stage data pipeline that resolves NS records, maps nameservers to IPs, associates those IPs with ASNs, and then maps each ASN to its parent organization. From there, I analyzed how many domains are effectively "single-provider dependent" versus partially diversified.

I also looked at In-Bailiwick hosting — where domains host their own nameservers (e.g., ns1.example.com serving example.com). That pattern is becoming increasingly rare as more domains move to large third-party providers, reinforcing the broader consolidation trend.

Beyond the findings themselves, the project was an exercise in building reproducible internet-scale measurement tooling. The codebase is designed to be easily rerunnable, with modular scripts, configurable concurrency, and clear documentation so other researchers can extend the analysis. Outputs include visualizations, distribution tables, and time-series comparisons that highlight how quickly consolidation is accelerating.

The research methodology involved a four-step data collection pipeline: NS record resolution for the top 10k domains from the Tranco list, A record resolution for all discovered nameservers, ASN (Autonomous System Number) acquisition for each IP address, and AS Organization mapping for each ASN. Analysis included identifying domains that are "unreachable" (exclusively using one AS organization) versus "affected" (partially using one AS organization).

The project demonstrates advanced DNS querying techniques, data pipeline architecture for internet-scale measurements, network infrastructure research methods, and statistical analysis with visualization. Results include detailed graphics showing domain distribution across providers, percentage tables comparing 2022 vs. 2024 data, and consolidation trends that underscore the rapid centralization of critical internet infrastructure.