Skip to main content
Definition Website Computer and Technology

Rate Limiting

Subject: Computer and Technology noun /ˈreɪt ˌlɪm.ɪ.tɪŋ/ 1 definition

  1. Definition (noun)

    Entry #77 · Updated

    Rate Limiting is a traffic-control mechanism used in APIs, web services, and computer networks to cap the rate at which a client, user, or IP address can submit requests or consume resources over a given interval. By tracking request frequency against predefined thresholds, an application or intermediary gateway either delays, throttles, or rejects excess traffic—commonly returning an HTTP 429 Too Many Requests status with a Retry-After header. Its core purpose is to protect backend stability, mitigate brute-force and denial-of-service attacks, prevent cascading failures, and ensure fair resource sharing across consumers.

    Common Forms

    • Token Bucket: Refills a virtual bucket with tokens at a fixed rate up to a maximum capacity, consuming one token per request so brief traffic bursts are permitted while keeping the long-term average rate bounded.

    • Leaky Bucket: Queues incoming requests in a fixed-size buffer and drains them at a constant outflow rate, smoothing spiky traffic into a steady stream and dropping requests when the queue overflows.

    • Fixed Window Counter: Tracks request counts within discrete, aligned time intervals (such as per minute) and resets at each boundary, offering simple implementation at the cost of potential double-burst spikes near window edges.

    • Sliding Window: Evaluates request timestamps across a rolling time span—using either sorted request logs or weighted counters from the previous interval—to enforce smooth, accurate limits without boundary-spike artifacts.

    Example

    An authentication service configures an API gateway rate limit of 5 login attempts per minute per IP address using a sliding window counter in Redis. When a credential-stuffing script sends 40 POST /login requests in three seconds from a single IP, the gateway processes the first 5 attempts and immediately rejects the remaining 35 with HTTP 429 Too Many Requests, keeping database load flat for legitimate users.