AWS Builder Center
From Static Websites to Multi-AZ Architecture: What Building My First Real AWS Setup Actually Taught Me

From Static Websites to Multi-AZ Architecture: What Building My First Real AWS Setup Actually Taught Me

A beginner's breakdown of three real AWS builds — S3/CloudFront, event-driven Lambda, and a multi-AZ VPC — and the wrong assumptions each one exposed.

I am a computer science student specializing in cloud computing primarily in AWS
I'm a Computer Systems Engineering student, and like a lot of people learning cloud, I started exactly where you'd expect: hosting a static website on S3. It felt like a huge win. Then I kept going — CloudFront, Lambda, and eventually a full VPC with an Application Load Balancer and Auto Scaling — and each step broke a different assumption I didn't know I had.
This isn't a "10 AWS services you must know" listicle. It's what actually happened when I built these things myself, what confused me, and what I'd tell someone who's exactly where I was three months ago.

Stage 1: S3 + CloudFront — where "it's just a bucket" stops being true

Hosting a static site on S3 is the classic first project, and for good reason: you get a working deployment in under an hour. But the part nobody's tutorial emphasized enough for me was why you need CloudFront in front of it at all.
My assumption going in was that S3 was "the hosting" and CloudFront was an optional speed boost. What I actually learned:
  • S3 static website hosting serves over HTTP by default — no HTTPS unless you put a CDN in front of it. CloudFront isn't just for speed; it's how you get TLS on a static site without provisioning a server.
  • Bucket policies are not the same thing as IAM policies, and mixing them up is the single most common reason a "public" site returns Access Denied. One governs the bucket itself; the other governs a user or role's permissions. Getting this distinction straight probably saved me the most future debugging time of anything in this whole project.
  • Cache invalidation is a real cost and a real delay — pushing an update to your site and then wondering why nothing changed is a rite of passage.
The takeaway for a beginner: don't treat S3 hosting as "done" once the site loads. Ask why each piece is there. That question is what turned a checklist into actual understanding.

Stage 2: Event-driven Lambda — the mental model shift

Moving from a static site to a Lambda function triggered by S3 uploads was where cloud computing stopped feeling like "a server, but rented" and started feeling genuinely different from anything I'd built before.
The hard part wasn't the code. It was the mindset:
  • You stop thinking about servers and start thinking about events. An object lands in a bucket → that's the trigger → the function runs → it does one small thing → it stops. No process running in the background waiting for work.
  • IAM execution roles are the real gatekeeper. A Lambda function is only as capable as the role attached to it. I lost real time to a function that "should have worked" but silently failed because its role didn't have permission to write to the destination I wanted — no crash, just nothing happening, which is a much harder bug to spot than an error message.
  • Cold starts are a real, felt thing, not a theoretical concept from a slide deck, once you're testing your own function and watching that first invocation lag.
The takeaway: serverless isn't "less to learn," it's "different things to learn." Permissions and triggers replace provisioning and uptime as your main concerns.

Stage 3: The VPC with ALB and Auto Scaling — where it got genuinely humbling

This is the project that took the assumptions of the first two and put pressure on all of them at once. A multi-AZ VPC with a load balancer and Auto Scaling group is the point where AWS stops being a service catalog and starts being networking you're responsible for.
What actually challenged me:
  • Subnets aren't just "public" or "private" as a label — they're defined by their route tables. I could describe the public/private split in theory before I built this. I didn't actually understand it until I traced a request and realized a subnet is "public" only because its route table sends 0.0.0.0/0 to an Internet Gateway — nothing more, nothing magic.
  • Multi-AZ isn't a checkbox, it's a design decision with cost implications. Every AZ you span is redundancy you're paying for. Understanding why you'd accept that cost (fault tolerance beats a cheaper single point of failure) mattered more than knowing the setup steps.
  • The ALB and Auto Scaling group have to agree with each other, and when they don't — health checks failing, targets never registering — the error messages point you at the wrong layer half the time. Debugging this taught me to check health check configuration before assuming the application code was broken.
The takeaway: this is the project where "I followed the steps and it worked" turned into "I understand why each piece has to exist." That shift is the actual goal of building beyond tutorials.

What I'd tell someone starting exactly where I started

If you're a student or early-career developer looking at AWS's service catalog and feeling the pull to learn everything at once — don't. Here's the sequence that actually built understanding for me, in order of conceptual difficulty:
  1. Static hosting (S3 + CloudFront) — learn the difference between bucket policies and IAM policies. This distinction shows up everywhere later.
  1. One event-driven Lambda function — learn to read an execution role and predict what it can and can't do before you run the function.
  1. A small VPC with a load balancer — don't reach for Auto Scaling and multi-AZ until you can explain, without looking it up, why a subnet is public or private.
Each stage exposed a wrong assumption I didn't know I was carrying. That's the actual value of building instead of watching — tutorials show you the "how," but a wrong assumption only surfaces when something doesn't work and you have to find out why.
If you're earlier in this journey than I am: the confusion is the process working, not a sign you're behind.


I'm currently building out a portfolio around cloud/DevOps and AI-assisted operations tooling — if you're on a similar path, I'd genuinely like to hear what tripped you up first.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article