CAP Theorem - System Design

Added: 2026-05-06

[00:00:00]If you don't know this, you are terrible at system design. In a distributed system, the CAP theorem states you can only guarantee two of these three different things. Here's an example of consistency. When we have two different nodes, whether you read from node A or node B, it doesn't matter because they'll both return to you the exact same amount. Now, you may get inconsistent data because if you read from node A, you get this number, but if you read from node B, you will get a different number. Now, let's add a third node called node C. And let's say that it got the most recent update, and it now has 702 followers. These nodes are all able to talk to each other, but in the event of a partition, those connections essentially cut off. Well, B would probably tell A that this is the correct number. However, since C is cut off from these nodes, it can't inform them that 702 is currently the correct answer. This is where the CAP theorem comes into play. You have to be partition tolerant, otherwise your system just wouldn't exist. It would be dead. Do you choose to be consistent or to be available? If you choose consistency, well, then you'd actually have to turn the app offline sometimes because you might not always get the most up-to-date answer. Or, do you sacrifice consistency, making your app available, but possibly returning stale data? So, I hope this makes some sense now. You have to be partition tolerant, otherwise you don't have a system. If you choose consistency, you may have to turn your app offline so it doesn't serve stale data. But if you don't want to turn your app offline, you sacrifice consistency and allow it to serve some stale data.