Serverless
advancedDefinition
A way of running code in the cloud where you don't manage any servers yourself. You upload your code, and the cloud provider runs it on demand: automatically starting it up when needed and shutting it down when it's idle. You only pay for the seconds it's actually running.
In the wild
A small charity has a 'donate' button that emails them a thank-you note. They don't need a server running 24/7 for this. They just upload the email-sending code as a serverless function. It runs for half a second whenever someone donates, and costs almost nothing the rest of the time.
More from Architecture
Caching
Saving the result of slow work somewhere fast so the next person who needs it gets it instantly. The first time something is calculated or fetched it's slow; every time after that, the saved copy is handed out until it gets too old to trust.
CAP Theorem
A rule about systems that store data on many computers at once. It says you can have at most two of three things: everyone always sees the same data, the system always responds, and the system keeps working even when the computers can't talk to each other. You always have to give one up.
CDN (Content Delivery Network)
A worldwide network of computers that keep copies of a website's images, videos, and other files in many cities at once. When you visit the site, you're served from the nearest copy. So the page loads faster and the original server isn't overwhelmed.
Client-Server Architecture
The basic shape of most apps and websites. The client is the program in front of you, your browser, your phone app, and the server is a powerful computer somewhere else that does the heavy lifting and stores the data. The client asks for things, the server answers.
Event-Driven Architecture
A way of building software where different parts don't call each other directly. Instead, one part announces 'something happened' and any other part that cares can react. This keeps the pieces loosely connected, so adding new reactions is easy.
Eventual Consistency
When data is stored in many places at once, an update doesn't reach all of them at the exact same moment. For a brief time, different copies might disagree. But if you wait a beat, they all catch up and end up matching.