[System Design] Rate Limiting Algorithms

Photo by Alisa Anton on Unsplash

Read full article for free if you don’t have medium membership: https://liverungrow.medium.com/system-design-rate-limiting-algorithms-8a26892fb97f?sk=a3a37f345e80fd2503ba5b494f105cec

Fixed Window

A simple counter that resets every window (e.g. every 60 seconds).
If the limit is 4 requests per minute:

• Client can send up to 4 requests between 00:00 and 00:59.
• At 01:00, the counter resets to zero.

https://www.linkedin.com/pulse/rate-limiting-simplified-redis-antematter-e0hnf/

❌ ‘Burst at boundary’ does not handle bursts of traffic efficiently. Users can send a huge burst of request at the end of the first window, and the beginning of the second window. (A client could send 4 requests at 00:59, and another 4 at 01:00.) This could potentially overload the server.

Sliding Window

The sliding window algorithm is used to limit the number of requests a client can make over a specific period. It’s more flexible than a fixed window, and smoother than token bucket in many cases.

You define a rate limit, like 100 requests per 60 seconds.