A public API handles bursts of short requests that go quiet for hours at a time. The team wants new instances to appear within seconds during a spike, capacity to drop to zero when idle, and billing only for the executions and resources actually used. Which compute option is the most cost-effective?
Show answer and explanation
Answer: C
Functions on the Consumption or Flex Consumption plan scale out quickly with demand, can scale to zero during quiet periods, and bill per execution and resource use. That matches a bursty, short-lived request pattern. Dedicated App Service plans and always-on AKS node pools keep capacity billed while idle, and a fixed-minimum scale set pays for instances even when traffic is absent.
Why the other options are wrong
- A. AKS node pools keep nodes running and billed even with no traffic, so idle hours cost money that a per-execution model avoids.
- B. A dedicated App Service plan bills for reserved instances continuously regardless of request volume, which is wasteful for a workload idle for hours at a time.
- D. A fixed-minimum scale set keeps VMs running and charged during quiet periods, which conflicts with the goal of paying only for executions.
Exam tip. Consumption-plan Functions scale to zero and bill per execution, the most cost-effective option for spiky event-driven code.