Cloud Engineer · Step 5 of 8 · 7 min
Serverless
Know when serverless functions and services save effort, and where they surprise you.
Servers you never see
Serverless does not mean there are no servers. It means you do not manage them. You upload a function or a container, and the provider runs it when a request or event arrives, scales it, and charges you per request and per unit of run time.
Common examples are AWS Lambda, Azure Functions, and Google Cloud Run and Cloud Run functions. Many databases, queues and storage services are serverless in the same sense.
Where it shines
Serverless suits work that comes in bursts or on events: resize an image when it is uploaded, run a nightly report, handle a webhook, serve an API with uneven traffic. When nothing happens, you pay little or nothing, and generous free monthly allowances often cover learning projects.
What to watch
A cold start is the delay when a function has not run recently and must start up. Functions have time and memory limits. Costs can jump if something calls your function in a loop, or if an attacker floods a public endpoint.
Set limits on concurrency where the provider allows it, protect public endpoints, and keep the budget alerts from the last step switched on.
Try it
Design a serverless flow on paper
- 1.Pick an event-driven task, such as: when a photo is uploaded, make a thumbnail.
- 2.Ask a chatbot the prompt below.
- 3.Mark every place where a loop or a flood of requests could run up a bill, and write the guardrail for each.
Design a serverless way to [task] on [AWS, Azure or Google Cloud]. Name the services, draw the flow as numbered steps, explain what triggers each step, and list the limits and guardrails I should set so a bug cannot cause a large bill.