SAA-C03 Practice Question 48: Lambda Provisioned Concurrency
AWS Certified Solutions Architect – Associate (SAA-C03) · Test 5 · question 8 of 10
A customer-facing API backed by AWS Lambda behind API Gateway must meet a p99 latency target. The function is written in Java and takes about two seconds to initialise. Traffic is steady on weekdays with a sharp, predictable rise at 09:00. Users report slow responses at the start of the morning peak and occasionally during the day, and CloudWatch shows the slow invocations carry an initialisation duration.
What should the solutions architect configure?
Show answer and explanation
Answer:
An initialisation duration on the slow invocations identifies these as cold starts, which a two-second Java init makes expensive. Provisioned concurrency keeps a configured number of execution environments initialised and ready, so requests routed to them skip initialisation entirely. Because the morning rise is predictable, scheduling an increase shortly before 09:00 has warm environments in place when traffic arrives, and the level can be lowered afterwards so the feature is not paid for around the clock.
Option A can shorten initialisation somewhat, since CPU scales with memory, but it does not remove the cold start; it makes a two-second penalty into a smaller one that still lands on the p99.
Option B controls how much concurrency the function may consume and guarantees capacity, but every one of those concurrent environments still cold-starts on first use. Reserved concurrency is about capacity limits, not initialisation latency.
Option C hides the symptom from the timeout's perspective and makes the user experience worse. A longer timeout means slow requests take longer to fail rather than returning quickly.