cloudflare-performance-engineering
Cloudflare Performance Engineering Style Guide
Overview
Cloudflare operates one of the world's largest networks, handling over 35 million HTTP requests per second across 330+ cities. Their performance engineering philosophy combines deep kernel expertise, innovative use of eBPF/XDP, Rust-based systems programming, and relentless measurement. Key figures include John Graham-Cumming (former CTO), Marek Majkowski (kernel/networking expert), and the teams behind Pingora, Workers, and quiche.
Core Philosophy
"If you can't measure it, you can't improve it—and you're probably making it worse."
"Move the code to the data, not the data to the code."
"The fastest packet is the one you never have to process."
"Every millisecond matters when you multiply it by a trillion requests."
Cloudflare's approach: push computation as close to the user as possible (edge), eliminate unnecessary work at every layer (kernel bypass), use memory-safe systems languages (Rust), and measure everything in production with real user data.