Culvert-rating scoring has a 400 ms p99 latency budget. Average ModelLatency looks fine, but p99 stayed over budget for an hour and nobody was paged. What should they configure?
Select an answer to reveal the explanation.
Short Explanation
The budget is 400 ms p99, and the average looks fine while the tail sits over budget. Alarm on p99, not on the mean. A scale-out policy is not a substitute for paging when that percentile breaches.
Full Explanation
An operational error budget stated as p99 needs a CloudWatch alarm on that percentile, not on the average. A fine mean can hide a long tail. Polly is not that alarm, and a Domain 3 scale-out policy is not a substitute for paging when the budget is breached.