A CRDT primer: Why CRDTs are perfect for eventual consistency

Description

When it comes to distributed systems, coordination is slow. CRDTs are data structures that promise strong eventual consistency for highly available systems without the costs of coordination. If you’ve read about the theory behind state-based CRDTs, you may have run across the term “monotonic join semi-lattice." Despite the intimidating name, this concept (drawn from order theory) is built out of familiar elements.

John Mumm starts with these familiar elements of theory to build up to the complete mathematical concept. You’ll look at how state-based CRDTs work, relying on the lessons you’ve learned from order theory. John explores why you might want to use CRDTs in your own system by looking at an imaginary photo-sharing app called BirdWatch. Users of this app are much more interested in getting immediate feedback than in seeing the most up-to-date possible view of the entire system, which means you should consider eventual consistency. If you need to design your own CRDTs in practice, it’s helpful to understand the theory underlying them.

Once you understand why CRDTs are such a perfect fit for eventual consistency, you’ll implement a simple CRDT called a G-Counter and see that something just a little more sophisticated than a G-Counter could be a great fit for BirdWatch or similar apps. You’ll leave understanding why that dreaded “monotonic join semi-lattice” isn’t really that complex after all.