Vertical vs Horizontal Scaling

There are only two ways to handle more traffic. Vertical scaling means buying a bigger machine — more CPU, more RAM. Horizontal scaling means adding more machines and putting a load balancer (a box that spreads incoming requests across servers) in front of them.

The animation starts with vertical scaling: one server grows as traffic rises. Watch it hit the ceiling and crack — that ceiling is real. There is a biggest machine you can rent, and until then that one machine is a single point of failure. When it dies, everything dies.

Then the box splits into many identical small servers behind a load balancer, and the request dots redistribute evenly. Notice the one condition that makes this work: the servers are stateless, meaning no server remembers anything about a user between requests. All state lives in a shared database or cache, so any server can handle any request — which is exactly why the load balancer can send each dot anywhere.

One caution to carry into interviews: modern hardware is far more powerful than most candidates assume, so a single beefy box is often the right first answer. Watch how much traffic the single server absorbs before it cracks — do not distribute before you have to.