Vertical vs Horizontal Scaling

Scalability is a system's ability to handle growing load - more users, requests, or data - by adding resources, ideally with a proportional and predictable gain in capacity. A system scales well when roughly doubling its resources roughly doubles the work it can do, without a redesign.

Scaling comes in two fundamental shapes.

• Vertical scaling (scaling up) means making a single machine more powerful: adding CPU cores, RAM, faster disks, or a bigger network card. The application often needs no changes because it still runs as one process on one box.
• Horizontal scaling (scaling out) means adding more machines and spreading the work across them. This is how internet-scale systems grow, but it forces new questions: how requests are distributed, how state is shared, and how partial failures are handled.

Horizontal scaling usually relies on a load balancer to distribute traffic across nodes. On the data tier, growth is absorbed with database replication to scale reads and sharding and partitioning to scale writes and storage.

A closely related idea is statelessness: horizontally scaled application servers should keep no local session state, so any node can serve any request. Shared state moves to a shared datastore instead of living on one box.

Difference between vertical and horizontal scaling

Vertical scaling is the simplest first move: no code changes, no distribution problems, and often cheaper at small scale. But one machine has a hard ceiling and remains a single point of failure. Horizontal scaling removes that ceiling and adds redundancy, at the cost of coordination complexity - traffic distribution, data consistency, and partial failure all become your problem. Most real systems scale vertically until it hurts, then scale horizontally on the tier where load concentrates.

• Start every capacity discussion by asking which resource is the bottleneck: CPU, memory, disk, or network.
• Say "scale vertically first for simplicity, then horizontally for the tier under pressure" - it shows pragmatism.
• Call out statelessness explicitly; interviewers wait to hear it.
• Avoid vague phrases like "just add more servers" without explaining how traffic and state are distributed.