Capacity Planning Cheatsheet
Capacity planning is about making design choices defensible.
A minimal template
- Users and traffic
- DAU/MAU (if relevant)
- QPS average and peak factor
- Data size
- size per record/object
- storage growth per day
- Bandwidth
- read/write bandwidth at peak
- Latency budget
- p95/p99 target
- where time goes (network, compute, DB)
Useful rules of thumb
- You don’t need perfect math; you need the right order of magnitude.
- Tail latency gets worse under saturation; plan for headroom.
- Caches reduce load but introduce invalidation complexity.
Common mistakes
- ignoring peak factor
- ignoring read/write asymmetry
- no plan for hot keys/hot partitions