Distributed Systems

Distributed systems are what you get once one process and one database are no longer enough. This branch is about the choices that appear the moment state lives on more than one machine: where a record sits, what happens when a node disappears, and which guarantees you can still make to a caller who is waiting.

You should already be comfortable with a request crossing a process boundary. The Web branch, especially Backend, is the assumed background. Databases are the neighbour on the other side: isolation and indexing explain why a single node can look consistent, and this branch explains what changes when that assumption breaks. If those pages are still thin, the posts here name the idea anyway and link back when the earlier branch fills in.

The published notes move from placing data around a cluster, through the consistency choice you actually make during a partition, toward messaging so services can agree without sharing a database. Read them in that order if the topic is new. If you already run a cache or a queue, start at the post that matches the failure you have seen.

I write these while I am still operating systems that make these tradeoffs. The maturity badge is there so you can tell a working explanation from a note I expect to revise. Prerequisites are declared per post, which means a later page does not pretend the earlier mechanism is optional when it is not.