Horizontal vs Vertical Scaling: Why Stateless Apps Handle Traffic Spikes Better

A developer's side project built on a single Node.js EC2 instance collapsed under sudden traffic after a streamer shared its URL, pushing CPU to 99% and triggering out-of-memory errors. The experience prompted a deeper look at the two core scaling strategies: vertical scaling, which upgrades a single machine's resources but hits cost and capacity ceilings, and horizontal scaling, which distributes load across multiple identical instances. The key obstacle to horizontal scaling was the app's reliance on in-memory session storage, local caching, and disk writes, all of which broke when traffic was routed to new nodes. The solution required making the application stateless by offloading session and cache data to external services like Redis or a shared database. This architectural shift allows instances to be added or removed dynamically, eliminates single points of failure, and often reduces costs compared to running one oversized server.