Indexes in Distributed Systems

Everything so far assumed one machine. Once data is spread across many machines, the word index changes meaning. The interesting question stops being which structure to use and becomes which machine the index lives on.

This is the part of the topic that shows up in system design interviews, and the part most candidates have never thought about.

1. What Sharding Takes Away

Sharding splits one table across many machines. Each machine, called a shard, holds a subset of the rows