Designing a Rate Limiter: Token Bucket vs Sliding Window, and Making It Distributed

Short summary

A detailed comparison of rate limiting algorithms — token bucket, sliding window log, sliding window counter, and fixed window — with practical guidance on making them distributed using Redis. The key insight is that atomic read-modify-write via Redis Lua scripts prevents race conditions across servers. Real-world examples from Stripe and Cloudflare illustrate production patterns.

• •Token bucket is cheap and burst-friendly; sliding window counter offers smoother limits at slightly higher cost
• •Distributed rate limiting requires atomic operations — Redis Lua scripts ensure no race conditions across servers
• •Stripe uses token bucket in Redis; Cloudflare pushes limiting to the edge for early rejection

Generated with AI, which can make mistakes.