Horizontal and vertical scaling

Vertical scaling means giving a server more resources — more processor, more memory. Horizontal scaling means adding more servers and sharing the work between them, usually behind a load balancer.

Vertical scaling is far simpler and should generally be tried first: it needs no change to the application and a modern machine goes a long way. Its limits are that there is a largest available size, the price curve steepens, and one server remains one thing that can fail.

Horizontal scaling has no ceiling in principle, but it demands an application that can run as several copies — no state held locally, no assumption that consecutive requests reach the same process. Building that in from the start is cheap; retrofitting it under load is not.