AWS Builder Center
Early Stage System Design

Early Stage System Design

Designing scalable systems and architecture that survive production and traffic. Doing best practices and and standard ways of doing things, optimizing cost and resources.

Eruscent and Resumaxxing are live, but I realized this week that real traffic would completely crush our budget. The catch? The code isn't the problem.
Spring Boot and FastAPI can handle concurrent load just fine. The bottleneck isn't the architecture; it's the economics. Railway's Hobby plan can scale to meet real traffic, but it uses usage-based billing, so you pay per unit of CPU, RAM, and egress. My monthly bill would spike right along with my traffic. That realization sent me down an infrastructure rabbit hole. I'm sharing what I learned, not as "here's a tutorial" but as "here's my current understanding; feel free to correct me":
VPS vs. Managed: A VPS solves the cost-efficiency problem. You get predictable, dedicated hardware for a flat fee. The tradeoff is you configure everything yourself, unless you use something like Coolify to simplify it.
Docker: Packages your backend into a container so you can run it consistently anywhere. Kubernetes: I used to think K8s was mainly about rerouting traffic when something crashes (load balancing). Turns out, it's the full system that decides where containers run, restarts the ones that die, and scales things up or down automatically.
Horizontal vs. Vertical Scaling: Horizontal means adding more containers or servers. "Vertical" means making one server stronger (more CPU/RAM).
The part that clicked for me: you don't just pick one over the other based on preference. You look at what is actually bottlenecked. Server maxed out with room left in the container? Scale vertically. Need redundancy so one crash doesn't take the whole app down? Scale horizontally.
Where I landed for now: skip Kubernetes; it's overkill at our scale. Get a real VPS. Start vertical, and go horizontal when traffic actually forces it. As a 3rd-year CompEng student, I'm trying to build the habit of catching problems before they hit production rather than after. But I also know premature optimization is a trap. Senior engineers: am I overengineering this by moving to a VPS now, or is it the right move to secure predictable infrastructure costs before real traffic hits?
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