Intro

This article is the first part of my Distributed Systems series. I'm looking forward to writing about this as I'm learning about distributed systems at the moment.

Read replication

Read replication is a solution for distributed systems that are read-heavy, but don't write a lot of data.

We would have a leader database, and then follower databases. The goal is to remember a specific value that has been sent by the client and written by the leader database.

This means every time someone wants to read specific data, they can get it right away, the cached version of it, from one of the follower databases.

If they write new data, that will take some time to get updated for the follower database, leading to a consistency problem. This means that the user could likely end up getting the stale value. Also, take into account, that the follower databases don't get updated at the same time, hence there will be some time during the update when they aren't the same.

This solution was invented early in web history since web architects discovered that there were lots more reads than writes to web traffic.

Visualizations