Analysis and visualization of DNS resolution and web performance across five encrypted DNS protocols — Do53 (UDP/TCP), DoT, DoH, and DoQ — measured from 7 AWS vantage points worldwide.
This repository contains the Jupyter notebooks and data used to produce the results of my Master's thesis at TU Munich.
I designed and ran a large-scale measurement study comparing DNS protocol performance:
- Measured DNS resolution times for 100+ public resolvers across 5 protocols
- Deployed measurement infrastructure on 7 AWS EC2 regions (US East/West, EU Central, Africa South, Asia Pacific NE/SE, South America East)
- Analyzed web page load times (Top 12 Tranco sites) to quantify the real-world impact of DNS protocol choice on browsing performance
- Built a custom testbed to isolate DNS-related latency from other page load factors
- DoQ (DNS over QUIC) achieves the lowest resolution times in most scenarios, especially benefiting from 0-RTT connection resumption
- DoH and DoT add measurable overhead compared to unencrypted DNS, but the gap narrows with connection reuse
- Geographic distance to the resolver has a larger impact on performance than protocol choice
- Web page load times are only marginally affected by DNS protocol, since DNS resolution is a small fraction of total load time
| Notebook | Thesis Section | Description |
|---|---|---|
Performance.ipynb |
§5.1.1 – 5.1.4 | DNS resolution time analysis across all protocols, resolvers, and vantage points |
Performance-Top5.ipynb |
§5.1.5 | Focused comparison of the top 5 resolvers |
Meta.ipynb |
§5.1 (resolver list) | Resolver metadata and selection criteria for DNS measurements |
Web-Performance.ipynb |
§5.2.1 – 5.2.5 | Page load time analysis (Top 12 websites × 5 protocols × 7 regions) |
Web-Performance-Testbed.ipynb |
§5.2.6 | Controlled testbed measurements isolating DNS latency |
Meta-Web.ipynb |
§5.2 (resolver list) | Resolver metadata for web performance measurements |
Vantage-Points-Map.ipynb |
Fig. 4.1 | World map visualization of the 7 AWS vantage points |
Compare-Misc.ipynb |
— | Cross-vantage-point comparison (supplementary, not in thesis) |
- Analysis: pandas, NumPy, SQLite
- Visualization: matplotlib
- Network: pyasn (IP-to-ASN mapping)
- Infrastructure: AWS EC2 (7 regions), custom Go-based measurement tool and DNS proxy