Network Request Benchmark

Compares HTTP request performance across different implementations.

Usage

bun script.ts

Benchmarks

Implementation	Time (ms)	Protocol
C native (http DNS)	~90-150	HTTP
C native OpenSSL (IP)	~180-200	HTTPS
C curl HTTP/2	~170-180	HTTPS
C native OpenSSL (DNS)	~170-200	HTTPS
C Boost.Asio	~180-220	HTTPS
C libcurl	~180-250	HTTPS
C Network.framework	~170-350	HTTPS
Bun fetch	~180-300	HTTPS

ms
350 ┤                                  █
300 ┤                                  █
250 ┤                        █         █
200 ┤              █   █     █   █     █
150 ┤    █         █   █     █   █     █
100 ┤    █         █   █     █   █     █
 50 ┤    █         █   █     █   █     █
  0 ┼────█─────────█───█─────█───█─────█────
      C DNS  C IP  HTTP/2  OpenSSL Boost libcurl Network Bun

Files

• script.ts - Main benchmark script (compiles C/C++ and runs tests)
• c-curl.c - C implementation using libcurl (HTTPS)
• c-native.c - C implementation using raw sockets with DNS (HTTP)
• c-native-ip.c - C implementation using raw sockets + OpenSSL with direct IP (HTTPS)
• c-boost.cpp - C++ implementation using Boost.Asio (HTTPS)
• c-native-openssl.c - C implementation using raw sockets + OpenSSL with DNS (HTTPS)
• c-networkfw.c - C implementation using Apple's Network.framework (HTTPS)
• c-http2.c - C implementation using curl with HTTP/2 (HTTPS)
• networkfw - Compiled executable for Network.framework
• http2 - Compiled executable for HTTP/2 benchmark

Key Findings

• Native C with raw sockets is fastest for HTTP
• DNS lookup adds minimal overhead (~2ms)
• OpenSSL with direct IP is faster than with DNS
• HTTP/2 (curl) performs similarly to HTTP/1.1 for single requests
• libcurl and Network.framework perform similarly
• Bun's fetch is competitive with native implementations