Distributed systems are groups of computers that work together as one team to store and manage your data.

Multi-Computer Teams

This is called Consistency, which means every person who looks at the data sees the exact same, most recent update.

All Computers in Sync

This is called Availability, ensuring the system always responds to you quickly, even if some of its parts are struggling.

Always Getting an Answer

This is called Partition Tolerance, and it means the system can keep running even if the cables or signals between computers break.

Surviving Connection Issues

The CAP Theorem states that when a network problem happens, a system can only be perfectly in sync or always available, but not both.

The Impossible Choice

A network partition is like a roadblock that prevents two groups of computers from talking to each other and sharing updates.

When Connections Break

Sometimes a system provides 'stale' data, which is information that hasn't been updated yet because of a communication delay.

Dealing With Old Data

Some systems choose to stay online even if they can't sync, preferring to show slightly old information rather than nothing at all.

Staying Open for Business

Other systems choose to stop working temporarily if they can't guarantee that every computer has the exact same, correct information.

Stopping to Stay Accurate

Eventual consistency is a compromise where the system stays online and promises that all computers will eventually have the same data once they can talk again.

Catching Up Later

Banks prioritize being in sync over being always available, because it is better to wait than to let someone spend money they don't have.

The Banking Dilemma

Social media sites prioritize staying online over being perfectly in sync, because seeing a post a few seconds late is better than the app not loading.

The News Feed Choice

In the real world, connections between computers will always fail eventually, so designers must plan for these trade-offs from the start.

What is Catching Up Later?