# What is the difference between horizontal and vertical scaling?

Source: https://www.colosolutions.com/answers/horizontal-vs-vertical-scaling/
Updated: 2026-08-27
Reviewed by: James San Filippo

Vertical scaling means making one machine bigger — more CPU, RAM or storage in the same server. Horizontal scaling means adding more machines and spreading load across them. Vertical is simpler and has a ceiling; horizontal has no practical ceiling but requires the application to run on several nodes at once. Colo Solutions customers commonly scale vertically inside a cabinet, then horizontally across cabinets once one machine is no longer enough.

## Key figures

- **Vertical (scale up):** Add CPU, RAM or storage to an existing machine
- **Horizontal (scale out):** Add more machines and distribute load between them
- **Vertical ceiling:** The largest single machine you can buy
- **Horizontal requirement:** The application must tolerate running on multiple nodes
- **Single point of failure:** Vertical keeps one; horizontal can remove it

## The short version

**Vertical scaling makes one machine bigger. Horizontal scaling adds more
machines.**

Vertical — "scaling up" — means putting more CPU, memory or storage into a server
you already run. Horizontal — "scaling out" — means running several servers and
distributing work across them.

## The real trade-off

Most comparisons list advantages of each. The decision usually turns on two
things.

**Vertical is simpler, and it stops.** Nothing about your application has to
change: it is the same machine with more in it. But every machine has a largest
possible configuration, and you will eventually reach it. Until then you also
still have one machine — so its failure is still your outage.

**Horizontal has no practical ceiling, and it demands something from the
software.** Adding nodes can continue indefinitely, and once work is spread
across several machines the loss of one need not be an outage. But the
application has to tolerate running in more than one place at once — shared
session state, database writes and file storage are where this gets difficult,
and retrofitting it to something that assumed a single machine is real work.

## The order most people actually do it in

Vertical first, because it is cheap and requires no rewrite. Horizontal when
either the ceiling is close or the single point of failure has become
unacceptable.

That second reason is worth separating out. Teams often scale horizontally
believing they need the capacity, when what they actually need is **redundancy** —
and that is a better reason, because it is the one that survives a hardware
failure at 3am.

## How this looks in colocation

In a colocation cabinet the distinction is concrete rather than abstract.

Scaling **vertically** means opening the chassis: more RAM, more disks, a
denser server in the same rack units. Your footprint and your power draw may
barely change.

Scaling **horizontally** means more machines, which means more rack units, more
power and more network ports — and eventually more cabinets. This is where the
facility's limits start to matter: available power per cabinet, cooling, and
whether space is available beside your existing equipment rather than across the
hall.

That is worth planning before it is urgent. At Colo Solutions cabinets come as
one-third rack (12RU) or full rack (42RU), with private cages for deployments
that outgrow them — but contiguous space is easier to reserve early than to find
later.