
Cloud Armor rate limiting is the recommended way to shield backends behind an Application Load Balancer from request bursts. Rate limiting rules live in a Cloud Armor security policy attached to the load balancer's backend service, and they come in two actions that behave differently once a client goes over its limit. A throttle rule drops only the excess requests, while a rate-based ban keeps blocking the client for a set period. For the Professional Cloud Network Engineer exam, the choice between them usually rests on one requirement: whether a client that slows down should be served again right away.
A throttle rule counts requests against a threshold that you choose. The enforce-on-key setting decides what is counted. With enforce-on-key set to IP, Cloud Armor counts each client address separately instead of counting all traffic together, which is what turns a rate limit into a per-client quota.
Each request then gets one of two actions:
deny-429, which returns HTTP 429 Too Many Requests.Throttling only drops the extra requests. As soon as a client's measured rate falls back under the limit, its requests get the conform action again and it is served normally. No ban duration applies, so there is no period during which a well-behaved client stays locked out.
A rate-based ban also counts requests per key, and it can also return a 429. The difference is what happens after the threshold is crossed. Once a client goes over the limit, a rate-based ban keeps blocking that client for the full ban period that the rule sets, even after its request rate drops. The client is held out for the whole period regardless of how it behaves in the meantime.
The two actions suit different goals:
Both actions can use the same key and return the same status code, so neither of those settings distinguishes them. The presence of a ban period is the distinction.
Rate limiting rules are the Cloud Armor feature that counts requests per client. Load balancing settings work on a different question: the RATE balancing mode tells the load balancer how to spread traffic across backends based on capacity, and it does not limit individual clients. Preconfigured WAF rules also serve a different purpose, matching attack patterns such as SQL injection and cross-site scripting rather than counting requests.
On the Professional Cloud Network Engineer exam, we would read a rate limiting requirement for what should happen after a client slows down, then pick throttle or rate-based ban from that.
Our Professional Cloud Network Engineer practice exams cover Cloud Armor rate limiting alongside Cloud Armor security policies and load balancer backend configuration, with detailed explanations that drill these distinctions.