Design of a Rate Limiter

High-level design

This design places a rate-limiting decision service behind the web tier. The web server supplies a policy descriptor, receives an allow or deny decision, and forwards allowed work to an API server. The Envoy rate-limit service, originally developed at Lyft, provides this configuration example:

domain: messaging
descriptors:
  - key: message_type
    value: marketing
    rate_limit:
      unit: day
      requests_per_unit: 5

The configuration applies to a descriptor bucket in the messaging domain.

Here, unit is day, requests_per_unit is 5, and the descriptor selects marketing messages. The snippet does not include a user or client descriptor, so it does not establish a separate five-message allowance for each client. Per-client enforcement needs identity in the configured descriptor and the request sent to the limiter.