A controlled partial failure can prevent total collapse#
Demand exceeds safe capacity
Trying to queue all 50,000 can explode latency, trigger timeouts and retries, and make useful throughput fall. Admission control rejects excess or lower-priority work cheaply so critical requests retain capacity.
Options include tenant quotas, lower-priority task drops, disabling expensive features, sampling telemetry, or cached responses. HTTP 429 or 503 can communicate rejection when appropriate. Define fairness and prevent retrying clients from erasing the protection.
Related practice problems
Use the manifest-backed prompts below to apply this lesson in a complete design.
- Addressing cascading failuresGoogle SRE
- Reliability design patternsMicrosoft Azure Architecture Center
Reach this point to mark the lesson as read.