Skip to content
WritingArticle
  • #serverless
  • #aws
  • #lambda
  • #cloud
  • #students

What Is Serverless? A Practical Guide for Students and New Engineers

Pawan Joshi

6 min read

Share

"Serverless" is one of the most misunderstood words in cloud computing. There are still servers. You just don't manage them.

In this guide I'll explain what serverless actually means, show the building blocks on AWS, and be honest about when it's the right choice and when it isn't. I teach this to students at AWS Student Builder Group GEU, and this is the explanation that works best.

Serverless in one sentence

Serverless means you write the code and the cloud provider runs, scales, patches and bills it for you, per request, with nothing running when nobody uses it.

Four properties make something serverless:

  1. No servers to manage. No operating system updates, no capacity planning, no SSH.
  2. Scales automatically. From zero to thousands of requests without you changing anything.
  3. Pay for use. You pay per request or per unit of work, not for idle machines.
  4. Built-in availability. The service runs across multiple data centers by default.

How it compares

A server you run (EC2)Containers you run (ECS, EKS)Serverless (Lambda, DynamoDB)
You manageOS, runtime, scaling, patchesImages, cluster or tasks, scaling rulesOnly your code and configuration
Idle costFull price, 24/7You pay for running tasksClose to zero
ScalingYou configure itYou configure itAutomatic, per request
Startup timeMinutesSecondsMilliseconds to about a second
Long-running workUnlimitedUnlimitedLambda: up to 15 minutes per invocation

None of these is "best". They're tools for different jobs, and many real systems mix them.

The serverless building blocks on AWS

You rarely use one service alone. A typical serverless app is a few managed services connected by events.

JobAWS serviceWhat it does
Run codeAWS LambdaRuns a function in response to an event: an HTTP request, a file upload, a queue message, a schedule
Expose an APIAmazon API GatewayTurns HTTPS requests into Lambda invocations, with auth, throttling and custom domains
Store dataAmazon DynamoDBKey-value and document database with single-digit-millisecond reads and on-demand billing
Store filesAmazon S3Object storage for images, uploads, backups and static websites
Connect servicesAmazon EventBridgeAn event bus that routes events between services and on schedules
Buffer workAmazon SQSA queue that absorbs spikes and retries failed work
Coordinate stepsAWS Step FunctionsVisual workflows with retries, branches and waits between steps
Sign users inAmazon CognitoUser sign-up, sign-in and tokens
Deliver globallyAmazon CloudFrontCDN for your website and API

A real example: a student event portal

Here's a system I'd build serverless without thinking twice: a registration portal for a campus tech event.

  1. CloudFront and S3 serve the website.
  2. API Gateway receives the registration form.
  3. Lambda validates the data and writes it to DynamoDB.
  4. DynamoDB Streams or EventBridge emits a "registered" event.
  5. A second Lambda sends the confirmation email.
  6. On event day, traffic jumps from almost nothing to thousands of requests. Everything scales automatically, and the bill stays tiny because most days nobody uses it.

That spiky, mostly idle traffic is where serverless shines. A server sized for event day would sit idle and cost money the rest of the year.

When serverless is the right choice

  • Traffic is spiky or unpredictable. Event portals, hackathon tools, launches, webhooks.
  • Workloads are event-driven. "When a file is uploaded, resize it." "Every night, generate a report."
  • You're a small team. No one has time to patch servers on a weekend.
  • You want to ship fast. Managed services handle the hard parts: scaling, retries, availability.
  • Cost matters at low scale. Generous free tiers make side projects and student projects almost free.

When it isn't

Serverless isn't free of trade-offs. Watch for these:

  • Long-running jobs. Lambda stops after 15 minutes. Use AWS Fargate, AWS Batch or Step Functions for longer work.
  • Constant, very high traffic. At steady heavy load, reserved containers or instances can cost less per request.
  • Cold starts. A function that hasn't run recently takes extra time on its first request. It's usually small, and there are ways to reduce it. I cover them in AWS Lambda Cold Starts Explained.
  • Special hardware. GPUs or very large memory needs fit better elsewhere.
  • Learning new patterns. Event-driven design, idempotency and distributed tracing take practice.

Myths I hear from students

"Serverless means no backend." You still write backend code. You just don't run the machines it runs on.

"Serverless is only Lambda." Lambda is one piece. DynamoDB, S3, SQS, EventBridge and Step Functions are serverless too, and a good design leans on them so you write less code.

"It's always cheaper." It's usually cheaper for low or spiky traffic. Do the math for steady high traffic.

"You can't test it locally." Tools like AWS SAM and the AWS Toolkit for VS Code let you invoke functions locally and deploy test stacks in minutes.

How to start learning serverless

Here's the path I recommend to students:

  1. Learn the basics of the cloud. Understand regions, IAM roles and pricing. The AWS Certified Cloud Practitioner is a good first goal.
  2. Build one API end to end. Follow Build a Serverless REST API with AWS Lambda, API Gateway and DynamoDB. Deploy it, break it, read the logs.
  3. Learn Infrastructure as Code. Write your stack in AWS SAM or AWS CDK instead of clicking in the console. It's how real teams work.
  4. Go event-driven. Add a queue and an event bus. Read Event-Driven Serverless with EventBridge, SQS and Lambda.
  5. Model data for DynamoDB. Learn to design from access patterns in DynamoDB Single-Table Design for Beginners.
  6. Watch the bill and the logs. Set an AWS Budget, then learn CloudWatch Logs, metrics and X-Ray traces.
  7. Share what you build. Write about it, present it at a meetup, help a friend deploy their first function. That's how you grow from learner to builder.

Key takeaways

  • Serverless means you manage code, not machines, and you pay for use.
  • On AWS, Lambda runs your code, and services like API Gateway, DynamoDB, S3, SQS and EventBridge do the rest.
  • It's ideal for spiky, event-driven and small-team workloads, and less ideal for long jobs or constant heavy load.
  • The fastest way to learn is to build and deploy one small project end to end.

I'm working toward becoming a serverless community builder and sharing everything I learn along the way. Read my roadmap to becoming an AWS Serverless Hero, or say hello if you're learning serverless too.