This repo contains a small, reproducible discrete-event simulation of a single-server “video embedding” pipeline with stochastic arrivals, heavy-tailed service times, heterogeneous users, and optional abandonment.
- Create an environment and install deps:
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt- Run the experiment (writes everything to
./outputs/):
MPLBACKEND=Agg python run_experiment.py --seed 123 --replications 30outputs/metrics.csv: one row per (regime, mechanism, replication)outputs/summary_table.csv: aggregated means + 95% CIsoutputs/report.md: narrative summary (includes paste-ready paragraph)outputs/*.png: plots