Why estimate at all
Rough numbers reveal whether a design needs one box or a thousand. Capacity estimation, often called back of the envelope math, turns hand waving about scale into figures that justify caches, sharding, and replication.
What to estimate
- Traffic in requests per second from users times their activity.
- Storage from item size times the count, grown over time.
- Bandwidth from request size times request rate.
- Memory for what you want to keep hot in cache.
A simple method
Start from given numbers, like daily active users, and derive the rest with round figures.
- Per day to per second by dividing by roughly 100 thousand seconds.
- Read to write ratio to size caches and replicas.
- Peak over average by multiplying for bursts.
Keep it honest
Use round numbers and say your assumptions aloud. The point is the order of magnitude, not three decimal places. A figure that is off by a factor of two is fine, but missing a factor of a thousand changes the whole design.
Key idea
Do quick back of the envelope math on traffic, storage, and bandwidth so the scale of your design is grounded in numbers rather than guesses.