IaaS, PaaS, SaaS, containers and serverless

The stack and who manages what

Every application sits on layers: networking, storage, servers, virtualisation, operating system, runtime, application, data. Cloud models differ in how many layers the provider handles for you.

Model	You manage	Provider manages	Example
On premises	everything	nothing	your own data centre
IaaS (infrastructure)	OS, runtime, app, data	hardware, virtualisation, network	virtual machines, block storage
PaaS (platform)	app and data	OS, runtime, scaling	managed app platforms, managed databases
FaaS / serverless	function code	everything else, including scaling	event-triggered functions
SaaS (software)	your usage and settings	the whole application	email, CRM, collaboration tools

More managed means less work and less control. Less managed means more flexibility, and more to operate.

100%

Figure SDA-CORE-13-1 · Who manages what in each model · © forcsguy.com

Virtual machines and containers

• A virtual machine emulates hardware and runs a complete guest OS. Strong isolation, heavier, slower to start.
• A container packages an app and its dependencies and runs as an isolated process on the host's kernel. Lightweight, starts in seconds, and runs the same everywhere ("works on my machine" problems shrink).
• Orchestrators such as Kubernetes schedule containers across machines, restart failed ones, scale and roll out updates.

100%

Figure SDA-CORE-13-2 · Virtual machines and containers · © forcsguy.com

Serverless

You provide a function; the platform runs it on demand and bills per execution. Strengths: no servers to manage, scales to zero and up automatically, and you pay for use. Limits: cold starts (first request after idle is slower), maximum run time, limited local state, and harder local testing. Best for event-driven, spiky or glue workloads.

Choosing

• Spiky or tiny workloads, small team: serverless or PaaS.
• Steady high load or special needs: containers or VMs with more control, often cheaper at scale.
• Need maximum control, compliance or custom hardware: IaaS or on premises.
• Use managed databases and queues where you can: operating them well is hard.
• Beware lock-in: prefer open formats (containers, standard SQL) when portability matters.

More on this site