How do you control Tokio thread usage per service?

How do you build a system that handles 90 million requests per second? That’s the scale that Cloudflare operates at, processing roughly 25% of all internet traffic through their global network of 330+ edge locations.

In this episode, we talk to Kevin Guthrie and Edward Wang from Cloudflare about Pingora, their open-source Rust-based proxy that replaced nginx across their entire infrastructure. We’ll find out why they chose Rust for mission-critical systems handling such massive scale, the technical challenges of replacing battle-tested infrastructure, and the lessons learned from “oxidizing” one of the internet’s largest networks.

Cloudflare is a global network designed to make everything you connect to the Internet secure, private, fast, and reliable. Their network spans 330+ cities worldwide and handles approximately 25% of all internet traffic. Cloudflare provides a range of services including DDoS protection, CDN, DNS, and serverless computing—all built on infrastructure that processes billions of requests every day.

Kevin Guthrie is a Software Architect and Principal Distributed Systems Engineer at Cloudflare working on Pingora and the production services built upon it. He specializes in performance optimization at scale. Kevin has deep expertise in building high-performance systems and has contributed to open-source projects that power critical internet infrastructure.

Edward Wang is a Systems Engineer at Cloudflare who has been instrumental in developing Pingora, Cloudflare’s Rust-based HTTP proxy framework. He co-authored the announcement of Pingora’s open source release. Edward’s work focuses on performance optimization, security, and building developer-friendly APIs for network programming.