You already know what problems the cloud solves. Now let's name the three main ways to host your infrastructure. Understanding their differences will allow you to make good decisions (and understand why many companies choose one or the other).

On-premise (often abbreviated as on-prem) means that the servers are on your own premises and are yours. You buy the hardware, set it up, maintain it, and protect it.

Analogy: it's like owning a car. You buy it, pay for insurance, inspections, repairs, and the garage. In return, it's entirely yours and you do whatever you want with it.

When it makes sense:

• You have very strict legal requirements that require data not to leave your premises.
• You have already invested a lot in hardware and haven't amortized it yet.
• You need total control over every physical component (very specific cases).

Drawbacks: everything we saw in subchapter 1.2 (initial investment, maintenance, slowness, etc.).

In the cloud model (public cloud), a provider like AWS, Azure, or Google Cloud owns the datacenters. You rent the capacity you need and pay per use.

Analogy: it's like using a taxi or car sharing. You don't buy the car; you pay for each ride. You don't worry about insurance, gas, or repairs.

When it makes sense:

• New companies or projects that want to start quickly and cheaply.
• Variable workloads (traffic spikes).
• When you want to reach users in multiple countries without building datacenters.
• When you prefer to focus on your product, not on maintaining hardware.

Drawbacks:

• At very large scale and with constant use, it can be more expensive than having your own already amortized hardware.
• You depend on an external provider (risk of vendor lock-in, which we will see later).

The hybrid model combines on-premise and cloud, connecting them to work together.

Analogy: you have your own car for daily use (on-premise) but rent a van on weekends when you move (cloud for the peaks).

When it makes sense:

• A traditional company that already has its datacenter but wants to use the cloud for new projects or demand spikes.
• Sensitive data that must stay at home, but applications that benefit from the scale of the cloud.
• Gradual migrations: moving things to the cloud little by little, without a sudden jump.

Real example: A bank keeps its account database on-premise due to regulation, but runs its mobile app and website in the cloud to absorb usage spikes at the end of the month. Both worlds are connected by a secure private line.

When we say "cloud" in this book, we mean the public cloud: datacenters shared by many clients (although your resources are isolated). There is also the private cloud, which is basically cloud-like infrastructure but dedicated to a single organization (often on-premise). The hybrid model usually combines public cloud with on-premise or private cloud.

• On-premise: you own the hardware. Maximum control, maximum responsibility.
• Cloud: you rent capacity from a provider. Fast, flexible, pay-per-use.
• Hybrid: you combine both, useful in migrations and when there is data that must stay at home.
• There is no universally "better" option: it depends on the context (regulation, budget, workload type).

In the next subchapter, we will delve into the three cloud service models —IaaS, PaaS, and SaaS— that define how much work you delegate to the provider.