πŸ’¬
βœ‰οΈ
πŸ“Έ
▢️

Loading…

Capacity planning: turn users into requests, storage, and bandwidth

Capacity Planning

Capacity Planning

2 of 5

Requests Per Second

Every other number flows from this one. Get it wrong and everything else is wrong.

RPS is the single most important number in system design. It tells you how many servers to buy, how large to size your DB, and how much bandwidth to provision. It all starts here.

Formula

RPS = (DAU Γ— actions/day) Γ· 86,400

DAUDaily Active Users β€” who actually opens the app today
actions/dayHow many API calls per user per day (browse, post, like)
86,400Seconds in a day β€” never changes, always divide by this

This gives average RPS. Multiply by 3–5Γ— for peak traffic (evenings, events, launches).

Worked exampleIllustrative social app (25M DAU)
1
DAU500M users Γ— 5% daily active
=25M DAU

Only a fraction of registered users open the app each day

2
Actions/dayfeed (5Γ—) + posts (1Γ—) + likes (8Γ—)
=14 actions

Every action is at least one API call

3
Total calls25M Γ— 14
=350M/day

This is the total daily request volume

4
Average RPS350M Γ· 86,400
=~4,050 RPS

Traffic is NOT spread evenly β€” this is just the average

5
Peak RPS4,050 Γ— 3Γ— peak multiplier
=~12,150 RPS

Plan for THIS number β€” average capacity collapses at peak

6
Servers12,150 Γ· 1,000 RPS/server
=13 servers

Plus 30% headroom = provision 17 servers minimum

Takeaway: 13 servers for 500M registered users. Not 500. Not 5,000. The math matters β€” over-provisioning by 10Γ— costs a million dollars/year in wasted compute.