DHT Churn Experiment

Measure and visualize churn times for Pkarr records in the Mainline DHT.

🚀 Overview

1. Rust Experiment (main.rs)

   • Publish: Pushes records to the DHT.
   • Churn Tracking: Periodically checks record resolvability; logs churn times to churns.csv.

2. Python Analysis (analyze.py)

   • Reads churns.csv
   • Computes key metrics (mean lifetime, half-life, hourly survival probability)
   • Plots the observed survival curve with an exponential model fit (saved as survival_plot.png)

⚡️ Setup

Rust

• Install Rust.
• Build & Run:

   cargo build --release
   cargo run --release -- --num-records 100 --stop-fraction 0.9 --ttl-s 604800 --sleep-duration-ms 1000 --max-hours 12
   # or simply `cargo run` for defaults

Python

• Install dependencies:

   pip install -r requirements.txt

• Generate analysis and plot:

   python analyze.py

🔧 Configuration

Argument	Default	Description
--num-records	100	Total records to publish
--stop-fraction	0.9	Fraction of churned records to stop
--max-hours	12	Max hours to stop collecting data
--ttl-s	604800	TTL for each record (seconds)
--sleep-duration-ms	1000	Sleep duration between resolves (milliseconds)

📈 Output

• Churn Data: churns.csv
• Analysis & Visualization: survival_plot.png
  Additionally, key metrics (mean lifetime, half-life, hourly survival probability, and 95% CI) are printed to the terminal.

💡 Notes

• Results may vary due to network conditions.
• A pkarr record is considered non-resolvable upon the first occurrence of None.
• Non-churned records are logged with time_s = 0 (and treated as right-censored in analysis).

📜 License

MIT License. Contributions welcome!