Ferron3 Bench Setup

Web server performance benchmarking setup for comparing multiple HTTP servers across realistic scenarios.

Structure

client/                         Load generation client
  benchmark.py                  h2load wrapper that sweeps concurrency levels and writes CSV results
server/                         Server provisioning and configuration
  bench                         CLI entry point for building, configuring, and starting servers
  lib/                          Shell libraries (build, config, certs, docker, etc.)
  scenarios/                    Per-scenario config templates (1-5)
  backends/                     Node.js backend servers for reverse-proxy scenarios
  payloads/                     Test file generation (10KB + 2MB)
  wp/                           WordPress bootstrap for scenario 3
  test/                         Docker-based automated validation

Quick Start

1. Set up the server

cd server

# Install build dependencies (Ubuntu)
sudo apt-get install -y build-essential autoconf automake libtool cmake pkg-config clang \
    libpcre2-dev zlib1g-dev libssl-dev libzstd-dev golang-go curl wget jq

# For Rust servers
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
source ~/.cargo/env

# Build a server (e.g. ferron3)
bench build ferron3

# Generate TLS certs and test payloads
bench generate-certs
bench generate-payloads

# Start the server
bench start scenario1-zstd ferron3

2. Run the benchmark client

cd client

# Sweep concurrency from 1 to 256 in steps of 32
python benchmark.py https://localhost 256 32 results.csv

See server/README.md for the full CLI reference, supported servers, scenarios, and troubleshooting.

Supported Servers

Server	Language	Scenarios
NGINX	C	1, 2, 3, 4, 5
Apache HTTP Server	C	1, 3
Caddy	Go	1, 2, 4, 5
HAProxy	C	2
Traefik	Go	2
Ferron 2	Rust	1, 2, 4
Ferron 3	Rust	1, 2, 3, 4, 5
OpenLiteSpeed	C/C++	3

Supported Scenarios

Scenario	Description
scenario1-zstd	Static file serving with Zstd compression
scenario2-reverse-proxy	Reverse proxy and load balancing to 4 Node.js backends
scenario3-wordpress	Cached WordPress with PHP-FPM
scenario4-observability	Observability overhead (the "telemetry tax")
scenario5-http3	HTTP/3 (QUIC) — experimental