How yagikota/danewebperf works
Project Type and Purpose
• DANEWebPerf is a containerized benchmarking tool (full-stack CLI + microservices style) written primarily in Go, augmented by Python dataset scripts.
• It orchestrates three Dockerized components—Firefox (with Selenium), Unbound (DNS cache), and Let’s DANE (DANE proxy)—to measure how DANE impacts web page load times under four scenarios.
• Results are run on EC2 and stored in an S3 bucket.
a. Orchestrator (EC2 host)
– Go CLI commands under cmd/ (dane-check, pageloadtime, etc.)
– Shell scripts (install.sh, prepare.sh, start.sh)
– Dataset preparers in dataset/
– Docker daemon and AWS CLI client
b. Docker Network (internal virtual network)
– Firefox Container
• Headless Firefox + Selenium (requirements.txt, pageload_measure.py)
– Unbound Container
• DNS cache server (Dockerfile, unbound.conf)
– Let’s DANE Container
• DANE-EE(3) proxy (Go code under docker/firefox/letsdane/)
c. External Services
– Internet (target websites; DNS root/authoritative servers)
– AWS S3 (results storage bucket)
d. Build Artifacts
– Docker images (built from docker/firefox/Dockerfile.firefox and docker/unbound/Dockerfile.*)
– Go binaries (built from cmd/* via Go modules go.mod/go.sum)
Relationships and Data Flows
• Orchestrator → Docker API: “docker build” and “docker run” commands (prepare.sh, start.sh)
• Docker Network internal traffic:
• Firefox → (direct) Unbound → public Internet → Unbound → Firefox
– Scenarios w/o DANE
• Firefox → Let’s DANE → Unbound → public Internet → Unbound → Let’s DANE → Firefox
– Scenarios w/ DANE
• Orchestrator → AWS S3: uploads HAR, PCAP, CSV results
• Dataset scripts → Orchestrator: domain lists and scraped HTML/CSV data
Architectural Patterns & Design Principles
• Microservices style: each major function isolated in its own container
• Clear separation of concerns:
• Measurement controller (Go CLI + shell scripts)
• Data collection agents (Firefox, Unbound, Let’s DANE)
• Dataset preparation (Python)
• Reproducible builds: Dockerfiles under docker/, Go modules, Python requirements