SAA-C03 Practice Question 75: API Gateway Usage Plan Throttling
AWS Certified Solutions Architect – Associate (SAA-C03) · Test 8 · question 5 of 10
A public REST API fronted by Amazon API Gateway is backed by a fixed-size database that starts timing out above roughly 600 requests per second. Most partners send modest traffic, but one partner's batch job occasionally floods the API and takes the service down for everyone. The company wants each partner held to its own agreed request rate and monthly call allowance, and wants partners that exceed it to receive a clear rejection rather than bringing the backend down.
What should the architect configure?
Show answer and explanation
Answer:
Usage plans are the API Gateway feature for per-client limits. Clients are identified by the API key they present, each key is associated with a plan, and the plan sets a throttling rate and burst using a token bucket, plus a quota for the number of calls over a period. Per-client limits are applied before per-method, account and Regional limits, so one partner's batch job is throttled on its own key and the others are unaffected. Requests over the limit get a 429 Too Many Requests response, which is exactly the clear rejection the company asked for, and the backend never sees the excess traffic.
Option A helps only where responses are cacheable and identical across partners. It does nothing to bound a partner that sends a flood of distinct requests, and the requirement is a per-partner allowance, which a cache has no concept of.
Option B targets the wrong layer and the wrong identity. API Gateway is a managed regional endpoint that you do not place behind your own load balancer, and a partner behind NAT presents one address for many callers while another partner may present many addresses.
Option C is the closest wrong answer. A WAF rate-based rule does limit requests, but it counts by source IP address rather than by client identity, so it cannot express "this partner's agreed rate", has no monthly quota concept, and blocks rather than returning the throttling response a partner's client is expected to back off on.
Reference: Throttle requests to your REST APIs for better throughput in API Gateway