Traffic to a language model is unpredictable and low. Which option fits best?
Why B is the answer
A serverless API or standard deployment
Pay-per-use suits variable traffic, because you are not paying for idle capacity.
In detail
- Pay-per-use suits low, unpredictable traffic.
- Move to provisioned only if traffic grows steady.
Use case
An illustrative scenario, not a real incident.
An internal tool gets a few requests a day. A serverless API or standard deployment charges for use, so idle time costs nothing.
See it in different ways
flowchart TD
A{"Traffic pattern"} -->|"Variable or low"| B["Serverless API or standard"]
A -->|"Steady and high"| C["Provisioned throughput"]mindmap
root((GenAIOps))
Prompts
In Git
Variants
Models
Fit on your data
Gradual rollout
Capacity
Standard
Provisioned
Evaluation
In CI/CD
On production tracesflowchart TD
Q{"What is traffic like?"}
Q -->|"Low and unpredictable"| A0["Serverless or standard"]
Q -->|"Steady and high"| A1["Provisioned"]Key point: Low, unpredictable traffic suits pay per use.
Why the other options are wrong
- A. Provisioned throughput, reserved for the whole year
- You pay for capacity that sits idle.
- C. A dedicated GPU cluster that runs at all times
- An always-on cluster wastes money.
- D. A batch job that runs once each night
- A nightly job cannot serve requests as they come.