Opening Scenario
Your team just shipped a feature: a PDF report generator. Users upload a request, your API triggers report generation, and the result is stored in S3. Currently this runs on a dedicated EC2 instance. The instance sits idle 95% of the time, but you still pay for it 24/7. When a product launch triggers 500 simultaneous requests, the single instance falls over. You spend a weekend provisioning more capacity, setting up auto-scaling groups, and tuning ECS task limits. There has to be a better model.
Question Progression
1. What if you didn't manage the server at all?
Question: The report generator runs for 3–30 seconds per request, then does nothing. Why does it need a persistent server?
Reasoning: A traditional server model keeps a process alive to accept the next request. But if requests are infrequent or bursty, you're paying for idle time and engineering around…
In the full tutorial
- When to Reach for Serverless