Cloudflare Rewrites Core System in Rust: CDN Performance Up 25%, Response Time Cut by 10ms

Cloudflare Rewrites Core System in Rust: CDN Performance Up 25%, Response Time Cut by 10ms

Cloudflare Rebuilds Core System FL in Rust — Achieves 25% Performance Boost

Date: 2025-11-10 13:26 Zhejiang

image

Cloudflare has rebuilt its core subsystem FL entirely in Rust, delivering a 25% performance increase alongside greater security and enhanced stability.

image

Key Announcement

Cloudflare revealed that its architectural migration to Rust shortened response time by 10 milliseconds and improved overall speed by 25%.

Rust delivered the dual advantage of high performance and increased safety, while also shortening development cycles.

---

From Pingora to FL2 — Reengineering “The Brain of Cloudflare”

After rewriting the Pingora subsystem, engineers tackled one of Cloudflare’s oldest and most critical systems — FL:

> FL is the central hub of Cloudflare.

> Upon receiving a request, FL applies security and performance functions across the network.

> It runs each customer’s configuration, WAF rules, DDoS protection, and traffic routing to developer platforms or Cloudflare’s R2 object storage service.

---

Building FL2 with Oxy

Cloudflare chose its in-house proxy framework Oxy for FL2 due to features such as:

  • Monitoring
  • Soft reloads
  • Dynamic configuration loading
  • Live switching

Highlight: Oxy’s graceful restart mechanism — Instead of closing all active connections when stopping a process, Oxy stops accepting new ones and serves existing sessions until they finish, ensuring no user disruption.

---

Overcoming a 15-Year Legacy

Challenges in replacing FL’s LuaJIT-based system after 15 years of continuous evolution:

  • Avoiding duplicate development for FL1 (old) and FL2 (new)
  • Creating a compatibility layer
  • → Allows Rust modules for FL2 to run within FL1
  • Enables teams to gradually replace Lua with Rust without waiting for full migration

---

Migration Strategy

Cloudflare’s engineers designed a multi-step rollout:

  • Testing via Flamingo
  • Sends thousands of end-to-end requests to FL1 and FL2 simultaneously
  • Phased traffic increases
  • Monitor metrics and performance at each stage
  • Fallback safety
  • If FL2 cannot handle a request, it is routed to FL1
  • Gradually increases FL2’s traffic share while reducing fallback rate

---

Performance Gains

Two primary drivers for FL2’s speed:

  • Written entirely in Rust, vs. FL1’s mix of C, Rust, and Lua
  • Eliminated language data conversion overhead

Result: FL2 uses half the CPU and less than half the memory compared to FL1.

---

Rust's Reliability Advantage

  • Compile-time safety prevents many runtime errors
  • Strict static checks and coding conventions
  • Thorough testing and code reviews
  • Most failures now linked to hardware, not software

---

Original link:

https://www.infoq.com/news/2025/10/cloudflare-rust-proxy/

---

Broader Context — Rust, AI, and Efficiency

This case shows how modern languages like Rust can transform performance, safety, and development speed in critical infrastructure.

Similar efficiency gains can be achieved in other fields — for example, AI-powered content publishing.

The open-source platform AiToEarn enables creators to:

  • Generate AI-driven content
  • Publish across multiple major platforms
  • Monetize efficiently
  • Access analytics and AI model rankings

Supported platforms: Douyin, Kwai, WeChat, Bilibili, Rednote, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, and X (Twitter).

Open-source repo: AiToEarn GitHub

---

AI Industry Perspectives

Leadership & Transition

Strategic Partnerships

Tech Decision Making

image

---

Further Reading

---

Conclusion

In today’s fast-changing AI and tech landscape, staying informed is not enough — understanding how strategy, leadership, and technology interact is key.

Platforms like AiToEarn demonstrate how AI innovation can pair with global content monetization, mirroring the agility and efficiency seen in Cloudflare’s Rust migration.

Creators can use such tools to align with technical progress and market dynamics, ensuring sustainable growth in a competitive digital world.

Read more

AI Coding Sprint "DeepSeek Moment": Gen Z Team Uses Domestic Model to Instantly Deliver Complex Apps, Surpassing Claude Code

AI Coding Sprint "DeepSeek Moment": Gen Z Team Uses Domestic Model to Instantly Deliver Complex Apps, Surpassing Claude Code

Cloud-Based AI Agents: Redefining the Programming Paradigm Cloud-based AI Agents are making significant advances, transforming how software is conceived, developed, and deployed. With zero human intervention, an “AI programming team” can directly deploy complex applications, leveraging ultra-large context capacities — reaching tens of millions in scale. Imagine simply stating your requirements,

By Honghao Wang