sysdesign-scalability-horizontal-vs-vertical

Horizontal scaling adds more machines, vertical scaling increases machine capacity, and partitioning divides data to manage capacity growth by classifying components as stateless, stateful local, or shared-write before implementing the architecture plan.

To scale a stateful service, classify it as stateful local or shared-write, evaluate the need to externalize state, and apply a partitioning strategy to distribute data and avoid write conflicts during capacity planning.

You need to externalize state when your component is classified as stateful shared-write, allowing horizontal scaling by moving data out of the service to prevent write conflicts and capacity bottlenecks.

Vertical scaling works for stateless services by increasing machine capacity, but horizontal scaling is typically preferred for stateless components to handle capacity growth more efficiently during architecture planning.

To document a scaling plan during an incident review, name the affected component, identify the golden signal or bottleneck, classify its state type, and record the chosen scaling axis to formalize the capacity adjustment.

The limitation of vertical scaling for shared-write components is hitting hardware capacity ceilings, requiring partitioning or state externalization to resolve bottlenecks and continue scaling architecture effectively.