An interactive explainer

One server dies and three quarters of your cache moves. It did not have to.

Picking a server with hash % N looks obvious and works fine — until N changes. Lose one machine out of four and nine of your twelve keys land somewhere new, even though only four of them were on the machine that died. A circle fixes it, and the fix costs something.

Consistent hashing · Sharding · Rebalancing

Systems · Intermediate · 10 min · How a cluster decides which server holds your data — an interactive explainer on ViLearn, with a simulation and a knowledge check. All explainers.

Related: What a queue really does when you are overloaded · Why a log beats a queue when two teams want the same event · How your operating system fakes an entire computer