PaaS vs. IaaS vs. Serverless: Which Model Fits?

Infrastructure as a service (IaaS) gives you virtual servers, networks and storage that you operate yourself. Platform as a service (PaaS) takes your code and runs it, including builds, deployments, scaling and certificates. Serverless goes one step further and runs individual functions or containers only when a request or event arrives, billing you per use. The difference is who runs which layer: the more the provider runs, the less you can and must control.

Who runs what

The classic way to compare the models is a stack of layers, a framing that goes back to the cloud definition published by the US standards institute NIST. The table shows who is typically responsible for each layer.

Layer	Own server	IaaS	PaaS	Serverless
Application code	You	You	You	You
Data and access rights	You	You	You	You
Runtime, frameworks	You	You	Provider	Provider
Scaling, load balancing	You	You (with provider tools)	Provider (you set limits)	Provider
Operating system	You	You	Provider	Provider
Virtualisation	You	Provider	Provider	Provider
Hardware, network, building	You	Provider	Provider	Provider

Two rows never move: your code and your data remain your responsibility in every model, including backups and who has access.

The three models in practice

IaaS: maximum control

With IaaS you rent virtual machines, block storage, networks and load balancers, from a single VPS to a full cloud account. You install the operating system packages, configure the web server, schedule updates and design for failure.

Typical uses: steady workloads where fixed capacity is cheaper than platform pricing; software with special requirements (custom kernels, unusual ports, GPU drivers); self-hosted platforms such as Coolify, which turn IaaS back into a private PaaS. See self-hosted PaaS.

Billing: per server and hour or month, plus storage and outbound traffic.

PaaS: focus on the application

A PaaS takes source code or a container image and runs it. You choose instance sizes or limits; the platform handles builds, rollouts, TLS, restarts and logs. Managed databases and caches are usually available as add-ons. A full explanation is in what a PaaS is.

Typical uses: web applications, APIs, background workers and scheduled jobs for small and medium teams that would rather ship features than maintain servers.

Billing: per instance and month, or by consumed CPU and memory, plus databases and traffic.

Serverless: pay per use

Serverless platforms start your code in response to an HTTP request, a queue message, a file upload or a timer, and stop it afterwards. Function-as-a-service products run individual functions; serverless container platforms run whole containers and scale them to zero when idle.

Typical uses: webhooks, image processing, scheduled tasks, APIs with very uneven traffic, glue between services.

Billing: per invocation and execution time, often with a free allowance.

Watch out for: cold starts (the first request after idle takes longer), execution time limits, and the temptation to split an application into dozens of functions that are hard to test and observe.

Comparison at a glance

	IaaS	PaaS	Serverless
Operating effort	High	Low	Very low
Control	High	Medium	Low
Cost with idle workload	Pays for idle capacity	Depends on model	Low
Cost with constant load	Often lowest	Medium	Can be high
Lock-in risk	Low	Medium	Higher (provider-specific triggers and APIs)
Time to first deploy	Hours to days	Minutes	Minutes

How to decide

Ask four questions in this order:

1. Does anyone on the team want to operate servers? If not, rule out IaaS for the main application.
2. Is the traffic steady or spiky? Steady load favours fixed capacity (IaaS or instance-priced PaaS); spiky or rare traffic favours serverless or usage-priced PaaS.
3. Are there special technical needs? Unusual networking, GPUs or long-running processes point to IaaS or containers.
4. What do customers and regulators expect? Data location and operator matter regardless of model; check region and provider seat.

For most small teams with a typical web application, the answer is a PaaS for the app, a managed database next to it, and serverless functions for a few event-driven jobs. Teams that already package everything as containers can also look at container deployment for small teams.

Don’t forget the bill

The models shift costs rather than remove them. IaaS saves on platform fees but costs admin time; PaaS charges for convenience; serverless is cheap until traffic becomes constant. Put your own numbers, including the hours you spend on operations, into the PaaS cost calculator before you commit.